update samples from Release-114 as a part of 1.38.0 SDK stable release

update samples from Release-122 as a part of  SDK release

CODE_OF_CONDUCT.md

@@ -0,0 +1,9 @@
+# Microsoft Open Source Code of Conduct
+
+This project has adopted the [Microsoft Open Source Code of Conduct](https://opensource.microsoft.com/codeofconduct/).
+
+Resources:
+
+- [Microsoft Open Source Code of Conduct](https://opensource.microsoft.com/codeofconduct/)
+- [Microsoft Code of Conduct FAQ](https://opensource.microsoft.com/codeofconduct/faq/)
+- Contact [opencode@microsoft.com](mailto:opencode@microsoft.com) with questions or concerns

README.md

@@ -1,77 +1,43 @@
-# Azure Machine Learning service example notebooks
+# Azure Machine Learning Python SDK notebooks
 
 > a community-driven repository of examples using mlflow for tracking can be found at https://github.com/Azure/azureml-examples
 
-This repository contains example notebooks demonstrating the [Azure Machine Learning](https://azure.microsoft.com/services/machine-learning-service/) Python SDK which allows you to build, train, deploy and manage machine learning solutions using Azure.  The AML SDK allows you the choice of using local or cloud compute resources, while managing and maintaining the complete data science workflow from the cloud.
+Welcome to the Azure Machine Learning Python SDK notebooks repository!
 
-![Azure ML Workflow](https://raw.githubusercontent.com/MicrosoftDocs/azure-docs/master/articles/machine-learning/media/concept-azure-machine-learning-architecture/workflow.png)
+## Getting started
 
+These notebooks are recommended for use in an Azure Machine Learning [Compute Instance](https://docs.microsoft.com/azure/machine-learning/concept-compute-instance), where you can run them without any additional set up.
 
-## Quick installation
-```sh
-pip install azureml-sdk
-```
-Read more detailed instructions on [how to set up your environment](./NBSETUP.md) using Azure Notebook service, your own Jupyter notebook server, or Docker.
+However, the notebooks can be run in any development environment with the correct `azureml` packages installed.
 
-## How to navigate and use the example notebooks?
-If you are using an Azure Machine Learning Notebook VM, you are all set. Otherwise, you should always run the [Configuration](./configuration.ipynb) notebook first when setting up a notebook library on a new machine or in a new environment. It configures your notebook library to connect to an Azure Machine Learning workspace, and sets up your workspace and compute to be used by many of the other examples. 
-This [index](./index.md) should assist in navigating the Azure Machine Learning notebook samples and encourage efficient retrieval of topics and content. 
-
-If you want to...
-
- * ...try out and explore Azure ML, start with image classification tutorials: [Part 1 (Training)](./tutorials/image-classification-mnist-data/img-classification-part1-training.ipynb) and [Part 2 (Deployment)](./tutorials/image-classification-mnist-data/img-classification-part2-deploy.ipynb).
- * ...learn about experimentation and tracking run history: [track and monitor experiments](./how-to-use-azureml/track-and-monitor-experiments).
- * ...train deep learning models at scale, first learn about [Machine Learning Compute](./how-to-use-azureml/training/train-on-amlcompute/train-on-amlcompute.ipynb), and then try [distributed hyperparameter tuning](./how-to-use-azureml/ml-frameworks/pytorch/train-hyperparameter-tune-deploy-with-pytorch/train-hyperparameter-tune-deploy-with-pytorch.ipynb) and [distributed training](./how-to-use-azureml/ml-frameworks/pytorch/distributed-pytorch-with-horovod/distributed-pytorch-with-horovod.ipynb).
- * ...deploy models as a realtime scoring service, first learn the basics by [deploying to Azure Container Instance](./how-to-use-azureml/deployment/deploy-to-cloud/model-register-and-deploy.ipynb), then learn how to [production deploy models on Azure Kubernetes Cluster](./how-to-use-azureml/deployment/production-deploy-to-aks/production-deploy-to-aks.ipynb).
- * ...deploy models as a batch scoring service: [create Machine Learning Compute for scoring compute](./how-to-use-azureml/training/train-on-amlcompute/train-on-amlcompute.ipynb) and [use Machine Learning Pipelines to deploy your model](https://aka.ms/pl-batch-scoring).
- * ...monitor your deployed models, learn about using [App Insights](./how-to-use-azureml/deployment/enable-app-insights-in-production-service/enable-app-insights-in-production-service.ipynb).
-
-## Tutorials
-
-The [Tutorials](./tutorials) folder contains notebooks for the tutorials described in the [Azure Machine Learning documentation](https://aka.ms/aml-docs).
-  
-## How to use Azure ML
-
-The [How to use Azure ML](./how-to-use-azureml) folder contains specific examples demonstrating the features of the Azure Machine Learning SDK
-
-- [Training](./how-to-use-azureml/training) - Examples of how to build models using Azure ML's logging and execution capabilities on local and remote compute targets
-- [Training with ML and DL frameworks](./how-to-use-azureml/ml-frameworks) - Examples demonstrating how to build and train machine learning models at scale on Azure ML and perform hyperparameter tuning.
-- [Manage Azure ML Service](./how-to-use-azureml/manage-azureml-service) - Examples how to perform tasks, such as authenticate against Azure ML service in different ways.
-- [Automated Machine Learning](./how-to-use-azureml/automated-machine-learning) - Examples using Automated Machine Learning to automatically generate optimal machine learning pipelines and models
-- [Machine Learning Pipelines](./how-to-use-azureml/machine-learning-pipelines) - Examples showing how to create and use reusable pipelines for training and batch scoring
-- [Deployment](./how-to-use-azureml/deployment) - Examples showing how to deploy and manage machine learning models and solutions
-- [Azure Databricks](./how-to-use-azureml/azure-databricks) - Examples showing how to use Azure ML with Azure Databricks
-- [Reinforcement Learning](./how-to-use-azureml/reinforcement-learning) - Examples showing how to train reinforcement learning agents
-
----
-## Documentation
-
- * Quickstarts, end-to-end tutorials, and how-tos on the [official documentation site for Azure Machine Learning service](https://docs.microsoft.com/en-us/azure/machine-learning/service/).
- * [Python SDK reference](https://docs.microsoft.com/en-us/python/api/overview/azure/ml/intro?view=azure-ml-py)
- * Azure ML Data Prep SDK [overview](https://aka.ms/data-prep-sdk), [Python SDK reference](https://aka.ms/aml-data-prep-apiref), and [tutorials and how-tos](https://aka.ms/aml-data-prep-notebooks).
-
----
-
-
-## Community Repository 
-Visit this [community repository](https://github.com/microsoft/MLOps/tree/master/examples) to find useful end-to-end sample notebooks. Also, please follow these [contribution guidelines](https://github.com/microsoft/MLOps/blob/master/contributing.md) when contributing to this repository.   
-
-## Projects using Azure Machine Learning
-
-Visit following repos to see projects contributed by Azure ML users:
- - [Learn about Natural Language Processing best practices using Azure Machine Learning service](https://github.com/microsoft/nlp)
- - [Pre-Train BERT models using Azure Machine Learning service](https://github.com/Microsoft/AzureML-BERT)
- - [Fashion MNIST with Azure ML SDK](https://github.com/amynic/azureml-sdk-fashion)
- - [UMass Amherst Student Samples](https://github.com/katiehouse3/microsoft-azure-ml-notebooks) - A number of end-to-end machine learning notebooks, including machine translation, image classification, and customer churn, created by students in the 696DS course at UMass Amherst.
- 
-## Data/Telemetry 
-This repository collects usage data and sends it to Microsoft to help improve our products and services. Read Microsoft's [privacy statement to learn more](https://privacy.microsoft.com/en-US/privacystatement)
-
-To opt out of tracking, please go to the raw markdown or .ipynb files and remove the following line of code:
+Install the `azureml.core` Python package:
 
 ```sh
-    "![Impressions](https://PixelServer20190423114238.azurewebsites.net/api/impressions/MachineLearningNotebooks/how-to-use-azureml/README.png)"
+pip install azureml-core
 ```
-This URL will be slightly different depending on the file. 
 
- ![Impressions](https://PixelServer20190423114238.azurewebsites.net/api/impressions/MachineLearningNotebooks/README.png)
+Install additional packages as needed:
+
+```sh
+pip install azureml-mlflow
+pip install azureml-dataset-runtime
+pip install azureml-automl-runtime
+pip install azureml-pipeline
+pip install azureml-pipeline-steps
+...
+```
+
+We recommend starting with one of the [quickstarts](tutorials/compute-instance-quickstarts).
+
+## Contributing
+
+This repository is a push-only mirror. Pull requests are ignored.
+
+## Code of Conduct
+
+This project has adopted the [Microsoft Open Source Code of Conduct](https://opensource.microsoft.com/codeofconduct/). Please see the [code of conduct](CODE_OF_CONDUCT.md) for details.
+
+## Reference
+
+- [Documentation](https://docs.microsoft.com/azure/machine-learning)
+

configuration.ipynb

@@ -103,7 +103,7 @@
       "source": [
         "import azureml.core\n",
         "\n",
-        "print(\"This notebook was created using version 1.23.0 of the Azure ML SDK\")\n",
+        "print(\"This notebook was created using version 1.38.0 of the Azure ML SDK\")\n",
         "print(\"You are currently using version\", azureml.core.VERSION, \"of the Azure ML SDK\")"
       ]
     },
@@ -254,6 +254,8 @@
         "\n",
         "Many of the sample notebooks use Azure ML managed compute (AmlCompute) to train models using a dynamically scalable pool of compute. In this section you will create default compute clusters for use by the other notebooks and any other operations you choose.\n",
         "\n",
+        "> Note that if you have an AzureML Data Scientist role, you will not have permission to create compute resources. Talk to your workspace or IT admin to create the compute targets described in this section, if they do not already exist.\n",
+        "\n",
         "To create a cluster, you need to specify a compute configuration that specifies the type of machine to be used and the scalability behaviors.  Then you choose a name for the cluster that is unique within the workspace that can be used to address the cluster later.\n",
         "\n",
         "The cluster parameters are:\n",

contrib/fairness/fairlearn-azureml-mitigation.ipynb

@@ -36,9 +36,9 @@
         "\n",
         "<a id=\"Introduction\"></a>\n",
         "## Introduction\n",
-        "This notebook shows how to use [Fairlearn (an open source fairness assessment and unfairness mitigation package)](http://fairlearn.github.io) and Azure Machine Learning Studio for a binary classification problem. This example uses the well-known adult census dataset. For the purposes of this notebook, we shall treat this as a loan decision problem. We will pretend that the label indicates whether or not each individual repaid a loan in the past. We will use the data to train a predictor to predict whether previously unseen individuals will repay a loan or not. The assumption is that the model predictions are used to decide whether an individual should be offered a loan. Its purpose is purely illustrative of a workflow including a fairness dashboard - in particular, we do **not** include a full discussion of the detailed issues which arise when considering fairness in machine learning. For such discussions, please [refer to the Fairlearn website](http://fairlearn.github.io/).\n",
+        "This notebook shows how to use [Fairlearn (an open source fairness assessment and unfairness mitigation package)](http://fairlearn.org) and Azure Machine Learning Studio for a binary classification problem. This example uses the well-known adult census dataset. For the purposes of this notebook, we shall treat this as a loan decision problem. We will pretend that the label indicates whether or not each individual repaid a loan in the past. We will use the data to train a predictor to predict whether previously unseen individuals will repay a loan or not. The assumption is that the model predictions are used to decide whether an individual should be offered a loan. Its purpose is purely illustrative of a workflow including a fairness dashboard - in particular, we do **not** include a full discussion of the detailed issues which arise when considering fairness in machine learning. For such discussions, please [refer to the Fairlearn website](http://fairlearn.org/).\n",
         "\n",
-        "We will apply the [grid search algorithm](https://fairlearn.github.io/master/api_reference/fairlearn.reductions.html#fairlearn.reductions.GridSearch) from the Fairlearn package using a specific notion of fairness called Demographic Parity. This produces a set of models, and we will view these in a dashboard both locally and in the Azure Machine Learning Studio.\n",
+        "We will apply the [grid search algorithm](https://fairlearn.org/v0.4.6/api_reference/fairlearn.reductions.html#fairlearn.reductions.GridSearch) from the Fairlearn package using a specific notion of fairness called Demographic Parity. This produces a set of models, and we will view these in a dashboard both locally and in the Azure Machine Learning Studio.\n",
         "\n",
         "### Setup\n",
         "\n",
@@ -46,9 +46,10 @@
         "Please see the [configuration notebook](../../configuration.ipynb) for information about creating one, if required.\n",
         "This notebook also requires the following packages:\n",
         "* `azureml-contrib-fairness`\n",
-        "* `fairlearn==0.4.6` (v0.5.0 will work with minor modifications)\n",
+        "* `fairlearn>=0.6.2` (pre-v0.5.0 will work with minor modifications)\n",
         "* `joblib`\n",
-        "* `shap`\n",
+        "* `liac-arff`\n",
+        "* `raiwidgets~=0.7.0`\n",
         "\n",
         "Fairlearn relies on features introduced in v0.22.1 of `scikit-learn`. If you have an older version already installed, please uncomment and run the following cell:"
       ]
@@ -85,10 +86,9 @@
       "outputs": [],
       "source": [
         "from fairlearn.reductions import GridSearch, DemographicParity, ErrorRate\n",
-        "from fairlearn.widget import FairlearnDashboard\n",
+        "from raiwidgets import FairnessDashboard\n",
         "\n",
         "from sklearn.compose import ColumnTransformer\n",
-        "from sklearn.datasets import fetch_openml\n",
         "from sklearn.impute import SimpleImputer\n",
         "from sklearn.linear_model import LogisticRegression\n",
         "from sklearn.model_selection import train_test_split\n",
@@ -112,9 +112,9 @@
       "metadata": {},
       "outputs": [],
       "source": [
-        "from fairness_nb_utils import fetch_openml_with_retries\n",
+        "from fairness_nb_utils import fetch_census_dataset\n",
         "\n",
-        "data = fetch_openml_with_retries(data_id=1590)\n",
+        "data = fetch_census_dataset()\n",
         "    \n",
         "# Extract the items we want\n",
         "X_raw = data.data\n",
@@ -137,7 +137,7 @@
       "outputs": [],
       "source": [
         "A = X_raw[['sex','race']]\n",
-        "X_raw = X_raw.drop(labels=['sex', 'race'],axis = 1)"
+        "X_raw = X_raw.drop(labels=['sex', 'race'], axis = 1)"
       ]
     },
     {
@@ -257,9 +257,9 @@
       "metadata": {},
       "outputs": [],
       "source": [
-        "FairlearnDashboard(sensitive_features=A_test, sensitive_feature_names=['Sex', 'Race'],\n",
-        "                   y_true=y_test,\n",
-        "                   y_pred={\"unmitigated\": unmitigated_predictor.predict(X_test)})"
+        "FairnessDashboard(sensitive_features=A_test,\n",
+        "                  y_true=y_test,\n",
+        "                  y_pred={\"unmitigated\": unmitigated_predictor.predict(X_test)})"
       ]
     },
     {
@@ -312,8 +312,8 @@
         "sweep.fit(X_train, y_train,\n",
         "          sensitive_features=A_train.sex)\n",
         "\n",
-        "# For Fairlearn v0.5.0, need sweep.predictors_\n",
-        "predictors = sweep._predictors"
+        "# For Fairlearn pre-v0.5.0, need sweep._predictors\n",
+        "predictors = sweep.predictors_"
       ]
     },
     {
@@ -330,16 +330,14 @@
       "outputs": [],
       "source": [
         "errors, disparities = [], []\n",
-        "for m in predictors:\n",
-        "    classifier = lambda X: m.predict(X)\n",
-        "    \n",
+        "for predictor in predictors:\n",
         "    error = ErrorRate()\n",
         "    error.load_data(X_train, pd.Series(y_train), sensitive_features=A_train.sex)\n",
         "    disparity = DemographicParity()\n",
         "    disparity.load_data(X_train, pd.Series(y_train), sensitive_features=A_train.sex)\n",
         "    \n",
-        "    errors.append(error.gamma(classifier)[0])\n",
-        "    disparities.append(disparity.gamma(classifier).max())\n",
+        "    errors.append(error.gamma(predictor.predict)[0])\n",
+        "    disparities.append(disparity.gamma(predictor.predict).max())\n",
         "    \n",
         "all_results = pd.DataFrame( {\"predictor\": predictors, \"error\": errors, \"disparity\": disparities})\n",
         "\n",
@@ -388,10 +386,9 @@
       "metadata": {},
       "outputs": [],
       "source": [
-        "FairlearnDashboard(sensitive_features=A_test, \n",
-        "                   sensitive_feature_names=['Sex', 'Race'],\n",
-        "                   y_true=y_test.tolist(),\n",
-        "                   y_pred=predictions_dominant)"
+        "FairnessDashboard(sensitive_features=A_test, \n",
+        "                  y_true=y_test.tolist(),\n",
+        "                  y_pred=predictions_dominant)"
       ]
     },
     {
@@ -410,7 +407,7 @@
         "<a id=\"AzureUpload\"></a>\n",
         "## Uploading a Fairness Dashboard to Azure\n",
         "\n",
-        "Uploading a fairness dashboard to Azure is a two stage process. The `FairlearnDashboard` invoked in the previous section relies on the underlying Python kernel to compute metrics on demand. This is obviously not available when the fairness dashboard is rendered in AzureML Studio. By default, the dashboard in Azure Machine Learning Studio also requires the models to be registered. The required stages are therefore:\n",
+        "Uploading a fairness dashboard to Azure is a two stage process. The `FairnessDashboard` invoked in the previous section relies on the underlying Python kernel to compute metrics on demand. This is obviously not available when the fairness dashboard is rendered in AzureML Studio. By default, the dashboard in Azure Machine Learning Studio also requires the models to be registered. The required stages are therefore:\n",
         "1. Register the dominant models\n",
         "1. Precompute all the required metrics\n",
         "1. Upload to Azure\n",
@@ -584,7 +581,7 @@
         "<a id=\"Conclusion\"></a>\n",
         "## Conclusion\n",
         "\n",
-        "In this notebook we have demonstrated how to use the `GridSearch` algorithm from Fairlearn to generate a collection of models, and then present them in the fairness dashboard in Azure Machine Learning Studio. Please remember that this notebook has not attempted to discuss the many considerations which should be part of any approach to unfairness mitigation. The [Fairlearn website](http://fairlearn.github.io/) provides that discussion"
+        "In this notebook we have demonstrated how to use the `GridSearch` algorithm from Fairlearn to generate a collection of models, and then present them in the fairness dashboard in Azure Machine Learning Studio. Please remember that this notebook has not attempted to discuss the many considerations which should be part of any approach to unfairness mitigation. The [Fairlearn website](http://fairlearn.org/) provides that discussion"
       ]
     },
     {

contrib/fairness/fairlearn-azureml-mitigation.yml

@@ -3,5 +3,7 @@ dependencies:
 - pip:
   - azureml-sdk
   - azureml-contrib-fairness
-  - fairlearn==0.4.6
+  - fairlearn>=0.6.2
   - joblib
+  - liac-arff
+  - raiwidgets~=0.16.0

contrib/fairness/fairness_nb_utils.py

@@ -4,7 +4,13 @@
 
 """Utilities for azureml-contrib-fairness notebooks."""
 
+import arff
+from collections import OrderedDict
+from contextlib import closing
+import gzip
+import pandas as pd
 from sklearn.datasets import fetch_openml
+from sklearn.utils import Bunch
 import time
 
 
@@ -15,7 +21,7 @@ def fetch_openml_with_retries(data_id, max_retries=4, retry_delay=60):
             print("Download attempt {0} of {1}".format(i + 1, max_retries))
             data = fetch_openml(data_id=data_id, as_frame=True)
             break
-        except Exception as e:
+        except Exception as e:  # noqa: B902
             print("Download attempt failed with exception:")
             print(e)
             if i + 1 != max_retries:
@@ -26,3 +32,80 @@ def fetch_openml_with_retries(data_id, max_retries=4, retry_delay=60):
         raise RuntimeError("Unable to download dataset from OpenML")
 
     return data
+
+
+_categorical_columns = [
+    'workclass',
+    'education',
+    'marital-status',
+    'occupation',
+    'relationship',
+    'race',
+    'sex',
+    'native-country'
+]
+
+
+def fetch_census_dataset():
+    """Fetch the Adult Census Dataset.
+
+    This uses a particular URL for the Adult Census dataset. The code
+    is a simplified version of fetch_openml() in sklearn.
+
+    The data are copied from:
+    https://openml.org/data/v1/download/1595261.gz
+    (as of 2021-03-31)
+    """
+    try:
+        from urllib import urlretrieve
+    except ImportError:
+        from urllib.request import urlretrieve
+
+    filename = "1595261.gz"
+    data_url = "https://rainotebookscdn.blob.core.windows.net/datasets/"
+
+    remaining_attempts = 5
+    sleep_duration = 10
+    while remaining_attempts > 0:
+        try:
+            urlretrieve(data_url + filename, filename)
+
+            http_stream = gzip.GzipFile(filename=filename, mode='rb')
+
+            with closing(http_stream):
+                def _stream_generator(response):
+                    for line in response:
+                        yield line.decode('utf-8')
+
+                stream = _stream_generator(http_stream)
+                data = arff.load(stream)
+        except Exception as exc:  # noqa: B902
+            remaining_attempts -= 1
+            print("Error downloading dataset from {} ({} attempt(s) remaining)"
+                  .format(data_url, remaining_attempts))
+            print(exc)
+            time.sleep(sleep_duration)
+            sleep_duration *= 2
+            continue
+        else:
+            # dataset successfully downloaded
+            break
+    else:
+        raise Exception("Could not retrieve dataset from {}.".format(data_url))
+
+    attributes = OrderedDict(data['attributes'])
+    arff_columns = list(attributes)
+
+    raw_df = pd.DataFrame(data=data['data'], columns=arff_columns)
+
+    target_column_name = 'class'
+    target = raw_df.pop(target_column_name)
+    for col_name in _categorical_columns:
+        dtype = pd.api.types.CategoricalDtype(attributes[col_name])
+        raw_df[col_name] = raw_df[col_name].astype(dtype, copy=False)
+
+    result = Bunch()
+    result.data = raw_df
+    result.target = target
+
+    return result

contrib/fairness/upload-fairness-dashboard.ipynb

@@ -30,7 +30,7 @@
         "1. [Training Models](#TrainingModels)\n",
         "1. [Logging in to AzureML](#LoginAzureML)\n",
         "1. [Registering the Models](#RegisterModels)\n",
-        "1. [Using the Fairlearn Dashboard](#LocalDashboard)\n",
+        "1. [Using the Fairness Dashboard](#LocalDashboard)\n",
         "1. [Uploading a Fairness Dashboard to Azure](#AzureUpload)\n",
         "    1. Computing Fairness Metrics\n",
         "    1. Uploading to Azure\n",
@@ -48,9 +48,10 @@
         "Please see the [configuration notebook](../../configuration.ipynb) for information about creating one, if required.\n",
         "This notebook also requires the following packages:\n",
         "* `azureml-contrib-fairness`\n",
-        "* `fairlearn==0.4.6` (should also work with v0.5.0)\n",
+        "* `fairlearn>=0.6.2` (also works for pre-v0.5.0 with slight modifications)\n",
         "* `joblib`\n",
-        "* `shap`\n",
+        "* `liac-arff`\n",
+        "* `raiwidgets~=0.7.0`\n",
         "\n",
         "Fairlearn relies on features introduced in v0.22.1 of `scikit-learn`. If you have an older version already installed, please uncomment and run the following cell:"
       ]
@@ -88,7 +89,6 @@
       "source": [
         "from sklearn import svm\n",
         "from sklearn.compose import ColumnTransformer\n",
-        "from sklearn.datasets import fetch_openml\n",
         "from sklearn.impute import SimpleImputer\n",
         "from sklearn.linear_model import LogisticRegression\n",
         "from sklearn.model_selection import train_test_split\n",
@@ -110,9 +110,9 @@
       "metadata": {},
       "outputs": [],
       "source": [
-        "from fairness_nb_utils import fetch_openml_with_retries\n",
+        "from fairness_nb_utils import fetch_census_dataset\n",
         "\n",
-        "data = fetch_openml_with_retries(data_id=1590)\n",
+        "data = fetch_census_dataset()\n",
         "    \n",
         "# Extract the items we want\n",
         "X_raw = data.data\n",
@@ -389,12 +389,11 @@
       "metadata": {},
       "outputs": [],
       "source": [
-        "from fairlearn.widget import FairlearnDashboard\n",
+        "from raiwidgets import FairnessDashboard\n",
         "\n",
-        "FairlearnDashboard(sensitive_features=A_test, \n",
-        "                   sensitive_feature_names=['Sex', 'Race'],\n",
-        "                   y_true=y_test.tolist(),\n",
-        "                   y_pred=ys_pred)"
+        "FairnessDashboard(sensitive_features=A_test, \n",
+        "                  y_true=y_test.tolist(),\n",
+        "                  y_pred=ys_pred)"
       ]
     },
     {
@@ -404,7 +403,7 @@
         "<a id=\"AzureUpload\"></a>\n",
         "## Uploading a Fairness Dashboard to Azure\n",
         "\n",
-        "Uploading a fairness dashboard to Azure is a two stage process. The `FairlearnDashboard` invoked in the previous section relies on the underlying Python kernel to compute metrics on demand. This is obviously not available when the fairness dashboard is rendered in AzureML Studio. The required stages are therefore:\n",
+        "Uploading a fairness dashboard to Azure is a two stage process. The `FairnessDashboard` invoked in the previous section relies on the underlying Python kernel to compute metrics on demand. This is obviously not available when the fairness dashboard is rendered in AzureML Studio. The required stages are therefore:\n",
         "1. Precompute all the required metrics\n",
         "1. Upload to Azure\n",
         "\n",

contrib/fairness/upload-fairness-dashboard.yml

@@ -3,5 +3,7 @@ dependencies:
 - pip:
   - azureml-sdk
   - azureml-contrib-fairness
-  - fairlearn==0.4.6
+  - fairlearn>=0.6.2
   - joblib
+  - liac-arff
+  - raiwidgets~=0.16.0

how-to-use-azureml/automated-machine-learning/automl_env.yml

@@ -2,9 +2,8 @@ name: azure_automl
 dependencies:
   # The python interpreter version.
   # Currently Azure ML only supports 3.5.2 and later.
-- pip==20.2.4
+- pip==21.1.2
 - python>=3.5.2,<3.8
-- nb_conda
 - boto3==1.15.18
 - matplotlib==2.1.0
 - numpy==1.18.5
@@ -18,12 +17,13 @@ dependencies:
 - holidays==0.9.11
 - pytorch::pytorch=1.4.0
 - cudatoolkit=10.1.243
+- tornado==6.1.0
 
 - pip:
   # Required packages for AzureML execution, history, and data preparation.
-  - azureml-widgets~=1.23.0
+  - azureml-widgets~=1.38.0
   - pytorch-transformers==1.0.0
   - spacy==2.1.8
   - https://aka.ms/automl-resources/packages/en_core_web_sm-2.1.0.tar.gz
-  - -r https://automlcesdkdataresources.blob.core.windows.net/validated-requirements/1.23.0/validated_win32_requirements.txt [--no-deps]
-  - PyJWT < 2.0.0
+  - -r https://automlsdkdataresources.blob.core.windows.net/validated-requirements/1.38.0/validated_win32_requirements.txt [--no-deps]
+  - arch==4.14

how-to-use-azureml/automated-machine-learning/automl_env_linux.yml

@@ -2,9 +2,8 @@ name: azure_automl
 dependencies:
   # The python interpreter version.
   # Currently Azure ML only supports 3.5.2 and later.
-- pip==20.2.4
+- pip==21.1.2
 - python>=3.5.2,<3.8
-- nb_conda
 - boto3==1.15.18
 - matplotlib==2.1.0
 - numpy==1.18.5
@@ -18,13 +17,13 @@ dependencies:
 - holidays==0.9.11
 - pytorch::pytorch=1.4.0
 - cudatoolkit=10.1.243
+- tornado==6.1.0
 
 - pip:
   # Required packages for AzureML execution, history, and data preparation.
-  - azureml-widgets~=1.23.0
+  - azureml-widgets~=1.38.0
   - pytorch-transformers==1.0.0
   - spacy==2.1.8
   - https://aka.ms/automl-resources/packages/en_core_web_sm-2.1.0.tar.gz
-  - -r https://automlcesdkdataresources.blob.core.windows.net/validated-requirements/1.23.0/validated_linux_requirements.txt [--no-deps]
-  - PyJWT < 2.0.0
-
+  - -r https://automlsdkdataresources.blob.core.windows.net/validated-requirements/1.38.0/validated_linux_requirements.txt [--no-deps]
+  - arch==4.14

how-to-use-azureml/automated-machine-learning/automl_env_mac.yml

@@ -2,10 +2,9 @@ name: azure_automl
 dependencies:
   # The python interpreter version.
   # Currently Azure ML only supports 3.5.2 and later.
-- pip==20.2.4
+- pip==21.1.2
 - nomkl
 - python>=3.5.2,<3.8
-- nb_conda
 - boto3==1.15.18
 - matplotlib==2.1.0
 - numpy==1.18.5
@@ -19,12 +18,13 @@ dependencies:
 - holidays==0.9.11
 - pytorch::pytorch=1.4.0
 - cudatoolkit=9.0
+- tornado==6.1.0
 
 - pip:
   # Required packages for AzureML execution, history, and data preparation.
-  - azureml-widgets~=1.23.0
+  - azureml-widgets~=1.38.0
   - pytorch-transformers==1.0.0
   - spacy==2.1.8
   - https://aka.ms/automl-resources/packages/en_core_web_sm-2.1.0.tar.gz
-  - -r https://automlcesdkdataresources.blob.core.windows.net/validated-requirements/1.23.0/validated_darwin_requirements.txt [--no-deps]
-  - PyJWT < 2.0.0
+  - -r https://automlsdkdataresources.blob.core.windows.net/validated-requirements/1.38.0/validated_darwin_requirements.txt [--no-deps]
+  - arch==4.14

how-to-use-azureml/automated-machine-learning/automl_setup_mac.sh

@@ -32,6 +32,7 @@ if [ $? -ne 0 ]; then
 fi
 
 sed -i '' 's/AZUREML-SDK-VERSION/latest/' $AUTOML_ENV_FILE
+brew install libomp
 
 if source activate $CONDA_ENV_NAME 2> /dev/null
 then

how-to-use-azureml/automated-machine-learning/check_conda_version.py

@@ -3,7 +3,7 @@ import platform
 
 try:
     import conda
-except:
+except Exception:
     print('Failed to import conda.')
     print('This setup is usually run from the base conda environment.')
     print('You can activate the base environment using the command "conda activate base"')

how-to-use-azureml/automated-machine-learning/classification-bank-marketing-all-features/auto-ml-classification-bank-marketing-all-features.yml

@@ -0,0 +1,4 @@
+name: auto-ml-classification-bank-marketing-all-features
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/classification-credit-card-fraud/auto-ml-classification-credit-card-fraud.ipynb

@@ -1,503 +1,483 @@
 {
-  "cells": [
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "Copyright (c) Microsoft Corporation. All rights reserved.\n",
-        "\n",
-        "Licensed under the MIT License."
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "![Impressions](https://PixelServer20190423114238.azurewebsites.net/api/impressions/MachineLearningNotebooks/how-to-use-azureml/automated-machine-learning/classification-credit-card-fraud/auto-ml-classification-credit-card-fraud.png)"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "# Automated Machine Learning\n",
-        "_**Classification of credit card fraudulent transactions on remote compute **_\n",
-        "\n",
-        "## Contents\n",
-        "1. [Introduction](#Introduction)\n",
-        "1. [Setup](#Setup)\n",
-        "1. [Train](#Train)\n",
-        "1. [Results](#Results)\n",
-        "1. [Test](#Test)\n",
-        "1. [Acknowledgements](#Acknowledgements)"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "## Introduction\n",
-        "\n",
-        "In this example we use the associated credit card dataset to showcase how you can use AutoML for a simple classification problem. The goal is to predict if a credit card transaction is considered a fraudulent charge.\n",
-        "\n",
-        "This notebook is using remote compute to train the model.\n",
-        "\n",
-        "If you are using an Azure Machine Learning Compute Instance, you are all set. Otherwise, go through the [configuration](../../../configuration.ipynb) notebook first if you haven't already to establish your connection to the AzureML Workspace. \n",
-        "\n",
-        "In this notebook you will learn how to:\n",
-        "1. Create an experiment using an existing workspace.\n",
-        "2. Configure AutoML using `AutoMLConfig`.\n",
-        "3. Train the model using remote compute.\n",
-        "4. Explore the results.\n",
-        "5. Test the fitted model."
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "## Setup\n",
-        "\n",
-        "As part of the setup you have already created an Azure ML `Workspace` object. For Automated ML you will need to create an `Experiment` object, which is a named object in a `Workspace` used to run experiments."
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "import logging\n",
-        "\n",
-        "from matplotlib import pyplot as plt\n",
-        "import pandas as pd\n",
-        "import os\n",
-        "\n",
-        "import azureml.core\n",
-        "from azureml.core.experiment import Experiment\n",
-        "from azureml.core.workspace import Workspace\n",
-        "from azureml.core.dataset import Dataset\n",
-        "from azureml.train.automl import AutoMLConfig"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "This sample notebook may use features that are not available in previous versions of the Azure ML SDK."
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "print(\"This notebook was created using version 1.23.0 of the Azure ML SDK\")\n",
-        "print(\"You are currently using version\", azureml.core.VERSION, \"of the Azure ML SDK\")"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "ws = Workspace.from_config()\n",
-        "\n",
-        "# choose a name for experiment\n",
-        "experiment_name = 'automl-classification-ccard-remote'\n",
-        "\n",
-        "experiment=Experiment(ws, experiment_name)\n",
-        "\n",
-        "output = {}\n",
-        "output['Subscription ID'] = ws.subscription_id\n",
-        "output['Workspace'] = ws.name\n",
-        "output['Resource Group'] = ws.resource_group\n",
-        "output['Location'] = ws.location\n",
-        "output['Experiment Name'] = experiment.name\n",
-        "pd.set_option('display.max_colwidth', -1)\n",
-        "outputDf = pd.DataFrame(data = output, index = [''])\n",
-        "outputDf.T"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "## Create or Attach existing AmlCompute\n",
-        "A compute target is required to execute the Automated ML run. In this tutorial, you create AmlCompute as your training compute resource.\n",
-        "#### Creation of AmlCompute takes approximately 5 minutes. \n",
-        "If the AmlCompute with that name is already in your workspace this code will skip the creation process.\n",
-        "As with other Azure services, there are limits on certain resources (e.g. AmlCompute) associated with the Azure Machine Learning service. Please read [this article](https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-manage-quotas) on the default limits and how to request more quota."
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "from azureml.core.compute import ComputeTarget, AmlCompute\n",
-        "from azureml.core.compute_target import ComputeTargetException\n",
-        "\n",
-        "# Choose a name for your CPU cluster\n",
-        "cpu_cluster_name = \"cpu-cluster-1\"\n",
-        "\n",
-        "# Verify that cluster does not exist already\n",
-        "try:\n",
-        "    compute_target = ComputeTarget(workspace=ws, name=cpu_cluster_name)\n",
-        "    print('Found existing cluster, use it.')\n",
-        "except ComputeTargetException:\n",
-        "    compute_config = AmlCompute.provisioning_configuration(vm_size='STANDARD_DS12_V2',\n",
-        "                                                           max_nodes=6)\n",
-        "    compute_target = ComputeTarget.create(ws, cpu_cluster_name, compute_config)\n",
-        "\n",
-        "compute_target.wait_for_completion(show_output=True)"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "# Data"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "### Load Data\n",
-        "\n",
-        "Load the credit card dataset from a csv file containing both training features and labels. The features are inputs to the model, while the training labels represent the expected output of the model. Next, we'll split the data using random_split and extract the training data for the model."
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "data = \"https://automlsamplenotebookdata.blob.core.windows.net/automl-sample-notebook-data/creditcard.csv\"\n",
-        "dataset = Dataset.Tabular.from_delimited_files(data)\n",
-        "training_data, validation_data = dataset.random_split(percentage=0.8, seed=223)\n",
-        "label_column_name = 'Class'"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "## Train\n",
-        "\n",
-        "Instantiate a AutoMLConfig object. This defines the settings and data used to run the experiment.\n",
-        "\n",
-        "|Property|Description|\n",
-        "|-|-|\n",
-        "|**task**|classification or regression|\n",
-        "|**primary_metric**|This is the metric that you want to optimize. Classification supports the following primary metrics: <br><i>accuracy</i><br><i>AUC_weighted</i><br><i>average_precision_score_weighted</i><br><i>norm_macro_recall</i><br><i>precision_score_weighted</i>|\n",
-        "|**enable_early_stopping**|Stop the run if the metric score is not showing improvement.|\n",
-        "|**n_cross_validations**|Number of cross validation splits.|\n",
-        "|**training_data**|Input dataset, containing both features and label column.|\n",
-        "|**label_column_name**|The name of the label column.|\n",
-        "\n",
-        "**_You can find more information about primary metrics_** [here](https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-configure-auto-train#primary-metric)"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "automl_settings = {\n",
-        "    \"n_cross_validations\": 3,\n",
-        "    \"primary_metric\": 'average_precision_score_weighted',\n",
-        "    \"enable_early_stopping\": True,\n",
-        "    \"max_concurrent_iterations\": 2, # This is a limit for testing purpose, please increase it as per cluster size\n",
-        "    \"experiment_timeout_hours\": 0.25, # This is a time limit for testing purposes, remove it for real use cases, this will drastically limit ablity to find the best model possible\n",
-        "    \"verbosity\": logging.INFO,\n",
-        "}\n",
-        "\n",
-        "automl_config = AutoMLConfig(task = 'classification',\n",
-        "                             debug_log = 'automl_errors.log',\n",
-        "                             compute_target = compute_target,\n",
-        "                             training_data = training_data,\n",
-        "                             label_column_name = label_column_name,\n",
-        "                             **automl_settings\n",
-        "                            )"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "Call the `submit` method on the experiment object and pass the run configuration. Depending on the data and the number of iterations this can run for a while. Validation errors and current status will be shown when setting `show_output=True` and the execution will be synchronous."
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "remote_run = experiment.submit(automl_config, show_output = False)"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "# If you need to retrieve a run that already started, use the following code\n",
-        "#from azureml.train.automl.run import AutoMLRun\n",
-        "#remote_run = AutoMLRun(experiment = experiment, run_id = '<replace with your run id>')"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "remote_run"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "## Results"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "#### Widget for Monitoring Runs\n",
-        "\n",
-        "The widget will first report a \"loading\" status while running the first iteration. After completing the first iteration, an auto-updating graph and table will be shown. The widget will refresh once per minute, so you should see the graph update as child runs complete.\n",
-        "\n",
-        "**Note:** The widget displays a link at the bottom. Use this link to open a web interface to explore the individual run details"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {
-        "tags": [
-          "widget-rundetails-sample"
-        ]
-      },
-      "outputs": [],
-      "source": [
-        "from azureml.widgets import RunDetails\n",
-        "RunDetails(remote_run).show()"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "remote_run.wait_for_completion(show_output=False)"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "#### Explain model\n",
-        "\n",
-        "Automated ML models can be explained and visualized using the SDK Explainability library. "
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "## Analyze results\n",
-        "\n",
-        "### Retrieve the Best Model\n",
-        "\n",
-        "Below we select the best pipeline from our iterations. The `get_output` method returns the best run and the fitted model.  Overloads on `get_output` allow you to retrieve the best run and fitted model for *any* logged metric or for a particular *iteration*."
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "best_run, fitted_model = remote_run.get_output()\n",
-        "fitted_model"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "#### Print the properties of the model\n",
-        "The fitted_model is a python object and you can read the different properties of the object.\n"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "## Test the fitted model\n",
-        "\n",
-        "Now that the model is trained, split the data in the same way the data was split for training (The difference here is the data is being split locally) and then run the test data through the trained model to get the predicted values."
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "# convert the test data to dataframe\n",
-        "X_test_df = validation_data.drop_columns(columns=[label_column_name]).to_pandas_dataframe()\n",
-        "y_test_df = validation_data.keep_columns(columns=[label_column_name], validate=True).to_pandas_dataframe()"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "# call the predict functions on the model\n",
-        "y_pred = fitted_model.predict(X_test_df)\n",
-        "y_pred"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "### Calculate metrics for the prediction\n",
-        "\n",
-        "Now visualize the data on a scatter plot to show what our truth (actual) values are compared to the predicted values \n",
-        "from the trained model that was returned."
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "from sklearn.metrics import confusion_matrix\n",
-        "import numpy as np\n",
-        "import itertools\n",
-        "\n",
-        "cf =confusion_matrix(y_test_df.values,y_pred)\n",
-        "plt.imshow(cf,cmap=plt.cm.Blues,interpolation='nearest')\n",
-        "plt.colorbar()\n",
-        "plt.title('Confusion Matrix')\n",
-        "plt.xlabel('Predicted')\n",
-        "plt.ylabel('Actual')\n",
-        "class_labels = ['False','True']\n",
-        "tick_marks = np.arange(len(class_labels))\n",
-        "plt.xticks(tick_marks,class_labels)\n",
-        "plt.yticks([-0.5,0,1,1.5],['','False','True',''])\n",
-        "# plotting text value inside cells\n",
-        "thresh = cf.max() / 2.\n",
-        "for i,j in itertools.product(range(cf.shape[0]),range(cf.shape[1])):\n",
-        "    plt.text(j,i,format(cf[i,j],'d'),horizontalalignment='center',color='white' if cf[i,j] >thresh else 'black')\n",
-        "plt.show()"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "## Acknowledgements"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "This Credit Card fraud Detection dataset is made available under the Open Database License: http://opendatacommons.org/licenses/odbl/1.0/. Any rights in individual contents of the database are licensed under the Database Contents License: http://opendatacommons.org/licenses/dbcl/1.0/ and is available at: https://www.kaggle.com/mlg-ulb/creditcardfraud\n",
-        "\n",
-        "The dataset has been collected and analysed during a research collaboration of Worldline and the Machine Learning Group (http://mlg.ulb.ac.be) of ULB (Universit\u00c3\u00a9 Libre de Bruxelles) on big data mining and fraud detection.\n",
-        "More details on current and past projects on related topics are available on https://www.researchgate.net/project/Fraud-detection-5 and the page of the DefeatFraud project\n",
-        "\n",
-        "Please cite the following works:\n",
-        "\n",
-        "Andrea Dal Pozzolo, Olivier Caelen, Reid A. Johnson and Gianluca Bontempi. Calibrating Probability with Undersampling for Unbalanced Classification. In Symposium on Computational Intelligence and Data Mining (CIDM), IEEE, 2015\n",
-        "\n",
-        "Dal Pozzolo, Andrea; Caelen, Olivier; Le Borgne, Yann-Ael; Waterschoot, Serge; Bontempi, Gianluca. Learned lessons in credit card fraud detection from a practitioner perspective, Expert systems with applications,41,10,4915-4928,2014, Pergamon\n",
-        "\n",
-        "Dal Pozzolo, Andrea; Boracchi, Giacomo; Caelen, Olivier; Alippi, Cesare; Bontempi, Gianluca. Credit card fraud detection: a realistic modeling and a novel learning strategy, IEEE transactions on neural networks and learning systems,29,8,3784-3797,2018,IEEE\n",
-        "\n",
-        "Dal Pozzolo, Andrea Adaptive Machine learning for credit card fraud detection ULB MLG PhD thesis (supervised by G. Bontempi)\n",
-        "\n",
-        "Carcillo, Fabrizio; Dal Pozzolo, Andrea; Le Borgne, Yann-A\u00c3\u00abl; Caelen, Olivier; Mazzer, Yannis; Bontempi, Gianluca. Scarff: a scalable framework for streaming credit card fraud detection with Spark, Information fusion,41, 182-194,2018,Elsevier\n",
-        "\n",
-        "Carcillo, Fabrizio; Le Borgne, Yann-A\u00c3\u00abl; Caelen, Olivier; Bontempi, Gianluca. Streaming active learning strategies for real-life credit card fraud detection: assessment and visualization, International Journal of Data Science and Analytics, 5,4,285-300,2018,Springer International Publishing\n",
-        "\n",
-        "Bertrand Lebichot, Yann-A\u00c3\u00abl Le Borgne, Liyun He, Frederic Obl\u00c3\u00a9, Gianluca Bontempi Deep-Learning Domain Adaptation Techniques for Credit Cards Fraud Detection, INNSBDDL 2019: Recent Advances in Big Data and Deep Learning, pp 78-88, 2019\n",
-        "\n",
-        "Fabrizio Carcillo, Yann-A\u00c3\u00abl Le Borgne, Olivier Caelen, Frederic Obl\u00c3\u00a9, Gianluca Bontempi Combining Unsupervised and Supervised Learning in Credit Card Fraud Detection Information Sciences, 2019"
-      ]
-    }
-  ],
-  "metadata": {
-    "authors": [
-      {
-        "name": "ratanase"
-      }
-    ],
-    "category": "tutorial",
-    "compute": [
-      "AML Compute"
-    ],
-    "datasets": [
-      "Creditcard"
-    ],
-    "deployment": [
-      "None"
-    ],
-    "exclude_from_index": false,
-    "file_extension": ".py",
-    "framework": [
-      "None"
-    ],
-    "friendly_name": "Classification of credit card fraudulent transactions using Automated ML",
-    "index_order": 5,
-    "kernelspec": {
-      "display_name": "Python 3.6",
-      "language": "python",
-      "name": "python36"
-    },
-    "language_info": {
-      "codemirror_mode": {
-        "name": "ipython",
-        "version": 3
-      },
-      "file_extension": ".py",
-      "mimetype": "text/x-python",
-      "name": "python",
-      "nbconvert_exporter": "python",
-      "pygments_lexer": "ipython3",
-      "version": "3.6.7"
-    },
-    "mimetype": "text/x-python",
-    "name": "python",
-    "nbconvert_exporter": "python",
-    "pygments_lexer": "ipython3",
-    "tags": [
-      "remote_run",
-      "AutomatedML"
-    ],
-    "task": "Classification",
-    "version": "3.6.7"
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Automated Machine Learning\n",
+    "_**Classification of credit card fraudulent transactions on remote compute **_\n",
+    "\n",
+    "## Contents\n",
+    "1. [Introduction](#Introduction)\n",
+    "1. [Setup](#Setup)\n",
+    "1. [Train](#Train)\n",
+    "1. [Results](#Results)\n",
+    "1. [Test](#Test)\n",
+    "1. [Acknowledgements](#Acknowledgements)"
+   ]
   },
-  "nbformat": 4,
-  "nbformat_minor": 2
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Introduction\n",
+    "\n",
+    "In this example we use the associated credit card dataset to showcase how you can use AutoML for a simple classification problem. The goal is to predict if a credit card transaction is considered a fraudulent charge.\n",
+    "\n",
+    "This notebook is using remote compute to train the model.\n",
+    "\n",
+    "If you are using an Azure Machine Learning Compute Instance, you are all set. Otherwise, go through the [configuration](../../../configuration.ipynb) notebook first if you haven't already to establish your connection to the AzureML Workspace. \n",
+    "\n",
+    "In this notebook you will learn how to:\n",
+    "1. Create an experiment using an existing workspace.\n",
+    "2. Configure AutoML using `AutoMLConfig`.\n",
+    "3. Train the model using remote compute.\n",
+    "4. Explore the results.\n",
+    "5. Test the fitted model."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Setup\n",
+    "\n",
+    "As part of the setup you have already created an Azure ML `Workspace` object. For Automated ML you will need to create an `Experiment` object, which is a named object in a `Workspace` used to run experiments."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import logging\n",
+    "\n",
+    "from matplotlib import pyplot as plt\n",
+    "import pandas as pd\n",
+    "import os\n",
+    "\n",
+    "import azureml.core\n",
+    "from azureml.core.experiment import Experiment\n",
+    "from azureml.core.workspace import Workspace\n",
+    "from azureml.core.dataset import Dataset\n",
+    "from azureml.train.automl import AutoMLConfig"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "This sample notebook may use features that are not available in previous versions of the Azure ML SDK."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "ws = Workspace.from_config()\n",
+    "\n",
+    "# choose a name for experiment\n",
+    "experiment_name = \"automl-classification-ccard-remote\"\n",
+    "\n",
+    "experiment = Experiment(ws, experiment_name)\n",
+    "\n",
+    "output = {}\n",
+    "output[\"Subscription ID\"] = ws.subscription_id\n",
+    "output[\"Workspace\"] = ws.name\n",
+    "output[\"Resource Group\"] = ws.resource_group\n",
+    "output[\"Location\"] = ws.location\n",
+    "output[\"Experiment Name\"] = experiment.name\n",
+    "pd.set_option(\"display.max_colwidth\", -1)\n",
+    "outputDf = pd.DataFrame(data=output, index=[\"\"])\n",
+    "outputDf.T"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Create or Attach existing AmlCompute\n",
+    "A compute target is required to execute the Automated ML run. In this tutorial, you create AmlCompute as your training compute resource.\n",
+    "\n",
+    "> Note that if you have an AzureML Data Scientist role, you will not have permission to create compute resources. Talk to your workspace or IT admin to create the compute targets described in this section, if they do not already exist.\n",
+    "\n",
+    "#### Creation of AmlCompute takes approximately 5 minutes. \n",
+    "If the AmlCompute with that name is already in your workspace this code will skip the creation process.\n",
+    "As with other Azure services, there are limits on certain resources (e.g. AmlCompute) associated with the Azure Machine Learning service. Please read [this article](https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-manage-quotas) on the default limits and how to request more quota."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.core.compute import ComputeTarget, AmlCompute\n",
+    "from azureml.core.compute_target import ComputeTargetException\n",
+    "\n",
+    "# Choose a name for your CPU cluster\n",
+    "cpu_cluster_name = \"cpu-cluster-1\"\n",
+    "\n",
+    "# Verify that cluster does not exist already\n",
+    "try:\n",
+    "    compute_target = ComputeTarget(workspace=ws, name=cpu_cluster_name)\n",
+    "    print(\"Found existing cluster, use it.\")\n",
+    "except ComputeTargetException:\n",
+    "    compute_config = AmlCompute.provisioning_configuration(\n",
+    "        vm_size=\"STANDARD_DS12_V2\", max_nodes=6\n",
+    "    )\n",
+    "    compute_target = ComputeTarget.create(ws, cpu_cluster_name, compute_config)\n",
+    "compute_target.wait_for_completion(show_output=True)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Data"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Load Data\n",
+    "\n",
+    "Load the credit card dataset from a csv file containing both training features and labels. The features are inputs to the model, while the training labels represent the expected output of the model. Next, we'll split the data using random_split and extract the training data for the model."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "data = \"https://automlsamplenotebookdata.blob.core.windows.net/automl-sample-notebook-data/creditcard.csv\"\n",
+    "dataset = Dataset.Tabular.from_delimited_files(data)\n",
+    "training_data, validation_data = dataset.random_split(percentage=0.8, seed=223)\n",
+    "label_column_name = \"Class\""
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Train\n",
+    "\n",
+    "Instantiate a AutoMLConfig object. This defines the settings and data used to run the experiment.\n",
+    "\n",
+    "|Property|Description|\n",
+    "|-|-|\n",
+    "|**task**|classification or regression|\n",
+    "|**primary_metric**|This is the metric that you want to optimize. Classification supports the following primary metrics: <br><i>accuracy</i><br><i>AUC_weighted</i><br><i>average_precision_score_weighted</i><br><i>norm_macro_recall</i><br><i>precision_score_weighted</i>|\n",
+    "|**enable_early_stopping**|Stop the run if the metric score is not showing improvement.|\n",
+    "|**n_cross_validations**|Number of cross validation splits.|\n",
+    "|**training_data**|Input dataset, containing both features and label column.|\n",
+    "|**label_column_name**|The name of the label column.|\n",
+    "\n",
+    "**_You can find more information about primary metrics_** [here](https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-configure-auto-train#primary-metric)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "automl_settings = {\n",
+    "    \"n_cross_validations\": 3,\n",
+    "    \"primary_metric\": \"average_precision_score_weighted\",\n",
+    "    \"enable_early_stopping\": True,\n",
+    "    \"max_concurrent_iterations\": 2,  # This is a limit for testing purpose, please increase it as per cluster size\n",
+    "    \"experiment_timeout_hours\": 0.25,  # This is a time limit for testing purposes, remove it for real use cases, this will drastically limit ablity to find the best model possible\n",
+    "    \"verbosity\": logging.INFO,\n",
+    "}\n",
+    "\n",
+    "automl_config = AutoMLConfig(\n",
+    "    task=\"classification\",\n",
+    "    debug_log=\"automl_errors.log\",\n",
+    "    compute_target=compute_target,\n",
+    "    training_data=training_data,\n",
+    "    label_column_name=label_column_name,\n",
+    "    **automl_settings,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Call the `submit` method on the experiment object and pass the run configuration. Depending on the data and the number of iterations this can run for a while. Validation errors and current status will be shown when setting `show_output=True` and the execution will be synchronous."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "remote_run = experiment.submit(automl_config, show_output=False)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# If you need to retrieve a run that already started, use the following code\n",
+    "# from azureml.train.automl.run import AutoMLRun\n",
+    "# remote_run = AutoMLRun(experiment = experiment, run_id = '<replace with your run id>')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Results"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Widget for Monitoring Runs\n",
+    "\n",
+    "The widget will first report a \"loading\" status while running the first iteration. After completing the first iteration, an auto-updating graph and table will be shown. The widget will refresh once per minute, so you should see the graph update as child runs complete.\n",
+    "\n",
+    "**Note:** The widget displays a link at the bottom. Use this link to open a web interface to explore the individual run details"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "tags": [
+     "widget-rundetails-sample"
+    ]
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.widgets import RunDetails\n",
+    "\n",
+    "RunDetails(remote_run).show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "remote_run.wait_for_completion(show_output=False)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Explain model\n",
+    "\n",
+    "Automated ML models can be explained and visualized using the SDK Explainability library. "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Analyze results\n",
+    "\n",
+    "### Retrieve the Best Model\n",
+    "\n",
+    "Below we select the best pipeline from our iterations. The `get_output` method returns the best run and the fitted model.  Overloads on `get_output` allow you to retrieve the best run and fitted model for *any* logged metric or for a particular *iteration*."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "best_run, fitted_model = remote_run.get_output()\n",
+    "fitted_model"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Print the properties of the model\n",
+    "The fitted_model is a python object and you can read the different properties of the object.\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Test the fitted model\n",
+    "\n",
+    "Now that the model is trained, split the data in the same way the data was split for training (The difference here is the data is being split locally) and then run the test data through the trained model to get the predicted values."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# convert the test data to dataframe\n",
+    "X_test_df = validation_data.drop_columns(\n",
+    "    columns=[label_column_name]\n",
+    ").to_pandas_dataframe()\n",
+    "y_test_df = validation_data.keep_columns(\n",
+    "    columns=[label_column_name], validate=True\n",
+    ").to_pandas_dataframe()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# call the predict functions on the model\n",
+    "y_pred = fitted_model.predict(X_test_df)\n",
+    "y_pred"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Calculate metrics for the prediction\n",
+    "\n",
+    "Now visualize the data on a scatter plot to show what our truth (actual) values are compared to the predicted values \n",
+    "from the trained model that was returned."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from sklearn.metrics import confusion_matrix\n",
+    "import numpy as np\n",
+    "import itertools\n",
+    "\n",
+    "cf = confusion_matrix(y_test_df.values, y_pred)\n",
+    "plt.imshow(cf, cmap=plt.cm.Blues, interpolation=\"nearest\")\n",
+    "plt.colorbar()\n",
+    "plt.title(\"Confusion Matrix\")\n",
+    "plt.xlabel(\"Predicted\")\n",
+    "plt.ylabel(\"Actual\")\n",
+    "class_labels = [\"False\", \"True\"]\n",
+    "tick_marks = np.arange(len(class_labels))\n",
+    "plt.xticks(tick_marks, class_labels)\n",
+    "plt.yticks([-0.5, 0, 1, 1.5], [\"\", \"False\", \"True\", \"\"])\n",
+    "# plotting text value inside cells\n",
+    "thresh = cf.max() / 2.0\n",
+    "for i, j in itertools.product(range(cf.shape[0]), range(cf.shape[1])):\n",
+    "    plt.text(\n",
+    "        j,\n",
+    "        i,\n",
+    "        format(cf[i, j], \"d\"),\n",
+    "        horizontalalignment=\"center\",\n",
+    "        color=\"white\" if cf[i, j] > thresh else \"black\",\n",
+    "    )\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Acknowledgements"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "This Credit Card fraud Detection dataset is made available under the Open Database License: http://opendatacommons.org/licenses/odbl/1.0/. Any rights in individual contents of the database are licensed under the Database Contents License: http://opendatacommons.org/licenses/dbcl/1.0/ and is available at: https://www.kaggle.com/mlg-ulb/creditcardfraud\n",
+    "\n",
+    "The dataset has been collected and analysed during a research collaboration of Worldline and the Machine Learning Group (http://mlg.ulb.ac.be) of ULB (Université Libre de Bruxelles) on big data mining and fraud detection.\n",
+    "More details on current and past projects on related topics are available on https://www.researchgate.net/project/Fraud-detection-5 and the page of the DefeatFraud project\n",
+    "\n",
+    "Please cite the following works:\n",
+    "\n",
+    "Andrea Dal Pozzolo, Olivier Caelen, Reid A. Johnson and Gianluca Bontempi. Calibrating Probability with Undersampling for Unbalanced Classification. In Symposium on Computational Intelligence and Data Mining (CIDM), IEEE, 2015\n",
+    "\n",
+    "Dal Pozzolo, Andrea; Caelen, Olivier; Le Borgne, Yann-Ael; Waterschoot, Serge; Bontempi, Gianluca. Learned lessons in credit card fraud detection from a practitioner perspective, Expert systems with applications,41,10,4915-4928,2014, Pergamon\n",
+    "\n",
+    "Dal Pozzolo, Andrea; Boracchi, Giacomo; Caelen, Olivier; Alippi, Cesare; Bontempi, Gianluca. Credit card fraud detection: a realistic modeling and a novel learning strategy, IEEE transactions on neural networks and learning systems,29,8,3784-3797,2018,IEEE\n",
+    "\n",
+    "Dal Pozzolo, Andrea Adaptive Machine learning for credit card fraud detection ULB MLG PhD thesis (supervised by G. Bontempi)\n",
+    "\n",
+    "Carcillo, Fabrizio; Dal Pozzolo, Andrea; Le Borgne, Yann-Aël; Caelen, Olivier; Mazzer, Yannis; Bontempi, Gianluca. Scarff: a scalable framework for streaming credit card fraud detection with Spark, Information fusion,41, 182-194,2018,Elsevier\n",
+    "\n",
+    "Carcillo, Fabrizio; Le Borgne, Yann-Aël; Caelen, Olivier; Bontempi, Gianluca. Streaming active learning strategies for real-life credit card fraud detection: assessment and visualization, International Journal of Data Science and Analytics, 5,4,285-300,2018,Springer International Publishing\n",
+    "\n",
+    "Bertrand Lebichot, Yann-Aël Le Borgne, Liyun He, Frederic Oblé, Gianluca Bontempi Deep-Learning Domain Adaptation Techniques for Credit Cards Fraud Detection, INNSBDDL 2019: Recent Advances in Big Data and Deep Learning, pp 78-88, 2019\n",
+    "\n",
+    "Fabrizio Carcillo, Yann-Aël Le Borgne, Olivier Caelen, Frederic Oblé, Gianluca Bontempi Combining Unsupervised and Supervised Learning in Credit Card Fraud Detection Information Sciences, 2019"
+   ]
+  }
+ ],
+ "metadata": {
+  "authors": [
+   {
+    "name": "ratanase"
+   }
+  ],
+  "category": "tutorial",
+  "compute": [
+   "AML Compute"
+  ],
+  "datasets": [
+   "Creditcard"
+  ],
+  "deployment": [
+   "None"
+  ],
+  "exclude_from_index": false,
+  "file_extension": ".py",
+  "framework": [
+   "None"
+  ],
+  "friendly_name": "Classification of credit card fraudulent transactions using Automated ML",
+  "index_order": 5,
+  "kernelspec": {
+   "display_name": "Python 3.6 - AzureML",
+   "language": "python",
+   "name": "python3-azureml"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.6.7"
+  },
+  "mimetype": "text/x-python",
+  "name": "python",
+  "nbconvert_exporter": "python",
+  "pygments_lexer": "ipython3",
+  "tags": [
+   "remote_run",
+   "AutomatedML"
+  ],
+  "task": "Classification",
+  "version": "3.6.7"
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
 }

how-to-use-azureml/automated-machine-learning/classification-credit-card-fraud/auto-ml-classification-credit-card-fraud.yml

@@ -0,0 +1,4 @@
+name: auto-ml-classification-credit-card-fraud
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/classification-text-dnn/auto-ml-classification-text-dnn.yml

@@ -0,0 +1,4 @@
+name: auto-ml-classification-text-dnn
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/classification-text-dnn/helper.py

@@ -4,53 +4,65 @@ from azureml.train.estimator import Estimator
 from azureml.core.run import Run
 
 
-def run_inference(test_experiment, compute_target, script_folder, train_run,
-                  train_dataset, test_dataset, target_column_name, model_name):
+def run_inference(
+    test_experiment,
+    compute_target,
+    script_folder,
+    train_run,
+    test_dataset,
+    target_column_name,
+    model_name,
+):
 
     inference_env = train_run.get_environment()
 
-    est = Estimator(source_directory=script_folder,
-                    entry_script='infer.py',
-                    script_params={
-                        '--target_column_name': target_column_name,
-                        '--model_name': model_name
-                    },
-                    inputs=[
-                        train_dataset.as_named_input('train_data'),
-                        test_dataset.as_named_input('test_data')
-                    ],
-                    compute_target=compute_target,
-                    environment_definition=inference_env)
+    est = Estimator(
+        source_directory=script_folder,
+        entry_script="infer.py",
+        script_params={
+            "--target_column_name": target_column_name,
+            "--model_name": model_name,
+        },
+        inputs=[test_dataset.as_named_input("test_data")],
+        compute_target=compute_target,
+        environment_definition=inference_env,
+    )
 
     run = test_experiment.submit(
-        est, tags={
-            'training_run_id': train_run.id,
-            'run_algorithm': train_run.properties['run_algorithm'],
-            'valid_score': train_run.properties['score'],
-            'primary_metric': train_run.properties['primary_metric']
-        })
+        est,
+        tags={
+            "training_run_id": train_run.id,
+            "run_algorithm": train_run.properties["run_algorithm"],
+            "valid_score": train_run.properties["score"],
+            "primary_metric": train_run.properties["primary_metric"],
+        },
+    )
 
-    run.log("run_algorithm", run.tags['run_algorithm'])
+    run.log("run_algorithm", run.tags["run_algorithm"])
     return run
 
 
 def get_result_df(remote_run):
 
     children = list(remote_run.get_children(recursive=True))
-    summary_df = pd.DataFrame(index=['run_id', 'run_algorithm',
-                                     'primary_metric', 'Score'])
+    summary_df = pd.DataFrame(
+        index=["run_id", "run_algorithm", "primary_metric", "Score"]
+    )
     goal_minimize = False
     for run in children:
-        if('run_algorithm' in run.properties and 'score' in run.properties):
-            summary_df[run.id] = [run.id, run.properties['run_algorithm'],
-                                  run.properties['primary_metric'],
-                                  float(run.properties['score'])]
-            if('goal' in run.properties):
-                goal_minimize = run.properties['goal'].split('_')[-1] == 'min'
+        if "run_algorithm" in run.properties and "score" in run.properties:
+            summary_df[run.id] = [
+                run.id,
+                run.properties["run_algorithm"],
+                run.properties["primary_metric"],
+                float(run.properties["score"]),
+            ]
+            if "goal" in run.properties:
+                goal_minimize = run.properties["goal"].split("_")[-1] == "min"
 
     summary_df = summary_df.T.sort_values(
-        'Score',
-        ascending=goal_minimize).drop_duplicates(['run_algorithm'])
-    summary_df = summary_df.set_index('run_algorithm')
+        "Score", ascending=goal_minimize
+    ).drop_duplicates(["run_algorithm"])
+    summary_df = summary_df.set_index("run_algorithm")
 
     return summary_df

how-to-use-azureml/automated-machine-learning/classification-text-dnn/infer.py

@@ -1,5 +1,6 @@
 import argparse
 
+import pandas as pd
 import numpy as np
 
 from sklearn.externals import joblib
@@ -11,19 +12,22 @@ from azureml.core.model import Model
 
 parser = argparse.ArgumentParser()
 parser.add_argument(
-    '--target_column_name', type=str, dest='target_column_name',
-    help='Target Column Name')
+    "--target_column_name",
+    type=str,
+    dest="target_column_name",
+    help="Target Column Name",
+)
 parser.add_argument(
-    '--model_name', type=str, dest='model_name',
-    help='Name of registered model')
+    "--model_name", type=str, dest="model_name", help="Name of registered model"
+)
 
 args = parser.parse_args()
 target_column_name = args.target_column_name
 model_name = args.model_name
 
-print('args passed are: ')
-print('Target column name: ', target_column_name)
-print('Name of registered model: ', model_name)
+print("args passed are: ")
+print("Target column name: ", target_column_name)
+print("Name of registered model: ", model_name)
 
 model_path = Model.get_model_path(model_name)
 # deserialize the model file back into a sklearn model
@@ -31,27 +35,31 @@ model = joblib.load(model_path)
 
 run = Run.get_context()
 # get input dataset by name
-test_dataset = run.input_datasets['test_data']
-train_dataset = run.input_datasets['train_data']
+test_dataset = run.input_datasets["test_data"]
 
-X_test_df = test_dataset.drop_columns(columns=[target_column_name]) \
-                        .to_pandas_dataframe()
-y_test_df = test_dataset.with_timestamp_columns(None) \
-                        .keep_columns(columns=[target_column_name]) \
-                        .to_pandas_dataframe()
-y_train_df = test_dataset.with_timestamp_columns(None) \
-                         .keep_columns(columns=[target_column_name]) \
-                         .to_pandas_dataframe()
+X_test_df = test_dataset.drop_columns(
+    columns=[target_column_name]
+).to_pandas_dataframe()
+y_test_df = (
+    test_dataset.with_timestamp_columns(None)
+    .keep_columns(columns=[target_column_name])
+    .to_pandas_dataframe()
+)
 
 predicted = model.predict_proba(X_test_df)
 
+if isinstance(predicted, pd.DataFrame):
+    predicted = predicted.values
+
 # Use the AutoML scoring module
-class_labels = np.unique(np.concatenate((y_train_df.values, y_test_df.values)))
 train_labels = model.classes_
+class_labels = np.unique(
+    np.concatenate((y_test_df.values, np.reshape(train_labels, (-1, 1))))
+)
 classification_metrics = list(constants.CLASSIFICATION_SCALAR_SET)
-scores = scoring.score_classification(y_test_df.values, predicted,
-                                      classification_metrics,
-                                      class_labels, train_labels)
+scores = scoring.score_classification(
+    y_test_df.values, predicted, classification_metrics, class_labels, train_labels
+)
 
 print("scores:")
 print(scores)

how-to-use-azureml/automated-machine-learning/continuous-retraining/auto-ml-continuous-retraining.yml

@@ -0,0 +1,4 @@
+name: auto-ml-continuous-retraining
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/continuous-retraining/register_model.py

@@ -25,9 +25,11 @@ datasets = [(Dataset.Scenario.TRAINING, train_ds)]
 
 # Register model with training dataset
 
-model = Model.register(workspace=ws,
-                       model_path=args.model_path,
-                       model_name=args.model_name,
-                       datasets=datasets)
+model = Model.register(
+    workspace=ws,
+    model_path=args.model_path,
+    model_name=args.model_name,
+    datasets=datasets,
+)
 
 print("Registered version {0} of model {1}".format(model.version, model.name))

how-to-use-azureml/automated-machine-learning/continuous-retraining/upload_weather_data.py

@@ -16,26 +16,82 @@ if type(run) == _OfflineRun:
 else:
     ws = run.experiment.workspace
 
-usaf_list = ['725724', '722149', '723090', '722159', '723910', '720279',
-             '725513', '725254', '726430', '720381', '723074', '726682',
-             '725486', '727883', '723177', '722075', '723086', '724053',
-             '725070', '722073', '726060', '725224', '725260', '724520',
-             '720305', '724020', '726510', '725126', '722523', '703333',
-             '722249', '722728', '725483', '722972', '724975', '742079',
-             '727468', '722193', '725624', '722030', '726380', '720309',
-             '722071', '720326', '725415', '724504', '725665', '725424',
-             '725066']
+usaf_list = [
+    "725724",
+    "722149",
+    "723090",
+    "722159",
+    "723910",
+    "720279",
+    "725513",
+    "725254",
+    "726430",
+    "720381",
+    "723074",
+    "726682",
+    "725486",
+    "727883",
+    "723177",
+    "722075",
+    "723086",
+    "724053",
+    "725070",
+    "722073",
+    "726060",
+    "725224",
+    "725260",
+    "724520",
+    "720305",
+    "724020",
+    "726510",
+    "725126",
+    "722523",
+    "703333",
+    "722249",
+    "722728",
+    "725483",
+    "722972",
+    "724975",
+    "742079",
+    "727468",
+    "722193",
+    "725624",
+    "722030",
+    "726380",
+    "720309",
+    "722071",
+    "720326",
+    "725415",
+    "724504",
+    "725665",
+    "725424",
+    "725066",
+]
 
 
 def get_noaa_data(start_time, end_time):
-    columns = ['usaf', 'wban', 'datetime', 'latitude', 'longitude', 'elevation',
-               'windAngle', 'windSpeed', 'temperature', 'stationName', 'p_k']
+    columns = [
+        "usaf",
+        "wban",
+        "datetime",
+        "latitude",
+        "longitude",
+        "elevation",
+        "windAngle",
+        "windSpeed",
+        "temperature",
+        "stationName",
+        "p_k",
+    ]
     isd = NoaaIsdWeather(start_time, end_time, cols=columns)
     noaa_df = isd.to_pandas_dataframe()
     df_filtered = noaa_df[noaa_df["usaf"].isin(usaf_list)]
     df_filtered.reset_index(drop=True)
-    print("Received {0} rows of training data between {1} and {2}".format(
-        df_filtered.shape[0], start_time, end_time))
+    print(
+        "Received {0} rows of training data between {1} and {2}".format(
+            df_filtered.shape[0], start_time, end_time
+        )
+    )
     return df_filtered
 
 
@@ -49,41 +105,53 @@ print("Argument 1(ds_name): %s" % args.ds_name)
 
 dstor = ws.get_default_datastore()
 register_dataset = False
+end_time = datetime.utcnow()
+
 try:
     ds = Dataset.get_by_name(ws, args.ds_name)
     end_time_last_slice = ds.data_changed_time.replace(tzinfo=None)
-    print("Dataset {0} last updated on {1}".format(args.ds_name,
-                                                   end_time_last_slice))
-except Exception as e:
+    print("Dataset {0} last updated on {1}".format(args.ds_name, end_time_last_slice))
+except Exception:
     print(traceback.format_exc())
-    print("Dataset with name {0} not found, registering new dataset.".format(args.ds_name))
+    print(
+        "Dataset with name {0} not found, registering new dataset.".format(args.ds_name)
+    )
     register_dataset = True
-    end_time_last_slice = datetime.today() - relativedelta(weeks=2)
+    end_time = datetime(2021, 5, 1, 0, 0)
+    end_time_last_slice = end_time - relativedelta(weeks=2)
 
-end_time = datetime.utcnow()
 train_df = get_noaa_data(end_time_last_slice, end_time)
 
 if train_df.size > 0:
-    print("Received {0} rows of new data after {0}.".format(
-        train_df.shape[0], end_time_last_slice))
-    folder_name = "{}/{:04d}/{:02d}/{:02d}/{:02d}/{:02d}/{:02d}".format(args.ds_name, end_time.year,
-                                                                        end_time.month, end_time.day,
-                                                                        end_time.hour, end_time.minute,
-                                                                        end_time.second)
+    print(
+        "Received {0} rows of new data after {1}.".format(
+            train_df.shape[0], end_time_last_slice
+        )
+    )
+    folder_name = "{}/{:04d}/{:02d}/{:02d}/{:02d}/{:02d}/{:02d}".format(
+        args.ds_name,
+        end_time.year,
+        end_time.month,
+        end_time.day,
+        end_time.hour,
+        end_time.minute,
+        end_time.second,
+    )
     file_path = "{0}/data.csv".format(folder_name)
 
     # Add a new partition to the registered dataset
     os.makedirs(folder_name, exist_ok=True)
     train_df.to_csv(file_path, index=False)
 
-    dstor.upload_files(files=[file_path],
-                       target_path=folder_name,
-                       overwrite=True,
-                       show_progress=True)
+    dstor.upload_files(
+        files=[file_path], target_path=folder_name, overwrite=True, show_progress=True
+    )
 else:
     print("No new data since {0}.".format(end_time_last_slice))
 
 if register_dataset:
-    ds = Dataset.Tabular.from_delimited_files(dstor.path("{}/**/*.csv".format(
-        args.ds_name)), partition_format='/{partition_date:yyyy/MM/dd/HH/mm/ss}/data.csv')
+    ds = Dataset.Tabular.from_delimited_files(
+        dstor.path("{}/**/*.csv".format(args.ds_name)),
+        partition_format="/{partition_date:yyyy/MM/dd/HH/mm/ss}/data.csv",
+    )
     ds.register(ws, name=args.ds_name)

how-to-use-azureml/automated-machine-learning/experimental/automl_setup_thin_client.cmd

@@ -5,7 +5,7 @@ set options=%3
 set PIP_NO_WARN_SCRIPT_LOCATION=0
 
 IF "%conda_env_name%"=="" SET conda_env_name="azure_automl_experimental"
-IF "%automl_env_file%"=="" SET automl_env_file="automl_env.yml"
+IF "%automl_env_file%"=="" SET automl_env_file="automl_thin_client_env.yml"
 
 IF NOT EXIST %automl_env_file% GOTO YmlMissing
 

how-to-use-azureml/automated-machine-learning/experimental/automl_setup_thin_client_linux.sh

@@ -12,7 +12,7 @@ fi
 
 if [ "$AUTOML_ENV_FILE" == "" ]
 then
-  AUTOML_ENV_FILE="automl_env.yml"
+  AUTOML_ENV_FILE="automl_thin_client_env.yml"
 fi
 
 if [ ! -f $AUTOML_ENV_FILE ]; then

how-to-use-azureml/automated-machine-learning/experimental/automl_setup_thin_client_mac.sh

@@ -12,7 +12,7 @@ fi
 
 if [ "$AUTOML_ENV_FILE" == "" ]
 then
-  AUTOML_ENV_FILE="automl_env.yml"
+  AUTOML_ENV_FILE="automl_thin_client_env_mac.yml"
 fi
 
 if [ ! -f $AUTOML_ENV_FILE ]; then

how-to-use-azureml/automated-machine-learning/experimental/automl_thin_client_env.yml

@@ -4,9 +4,10 @@ dependencies:
   # Currently Azure ML only supports 3.5.2 and later.
 - pip<=19.3.1
 - python>=3.5.2,<3.8
-- nb_conda
 - cython
 - urllib3<1.24
+- PyJWT < 2.0.0
+- numpy==1.18.5
 
 - pip:
   # Required packages for AzureML execution, history, and data preparation.
@@ -14,4 +15,3 @@ dependencies:
   - azureml-sdk
   - azureml-widgets
   - pandas
-  - PyJWT < 2.0.0

how-to-use-azureml/automated-machine-learning/experimental/automl_thin_client_env_mac.yml

@@ -5,9 +5,10 @@ dependencies:
 - pip<=19.3.1
 - nomkl
 - python>=3.5.2,<3.8
-- nb_conda
 - cython
 - urllib3<1.24
+- PyJWT < 2.0.0
+- numpy==1.18.5
 
 - pip:
   # Required packages for AzureML execution, history, and data preparation.
@@ -15,4 +16,3 @@ dependencies:
   - azureml-sdk
   - azureml-widgets
   - pandas
-  - PyJWT < 2.0.0

how-to-use-azureml/automated-machine-learning/experimental/classification-credit-card-fraud-local-managed/auto-ml-classification-credit-card-fraud-local-managed.ipynb

@@ -0,0 +1,420 @@
+{
+  "cells": [
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "Copyright (c) Microsoft Corporation. All rights reserved.\n",
+        "\n",
+        "Licensed under the MIT License."
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "![Impressions](https://PixelServer20190423114238.azurewebsites.net/api/impressions/MachineLearningNotebooks/how-to-use-azureml/automated-machine-learning/experimental/classification-credit-card-fraud/auto-ml-classification-credit-card-fraud.png)"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "# Automated Machine Learning\n",
+        "_**Classification of credit card fraudulent transactions on local managed compute **_\n",
+        "\n",
+        "## Contents\n",
+        "1. [Introduction](#Introduction)\n",
+        "1. [Setup](#Setup)\n",
+        "1. [Train](#Train)\n",
+        "1. [Results](#Results)\n",
+        "1. [Test](#Test)\n",
+        "1. [Acknowledgements](#Acknowledgements)"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "## Introduction\n",
+        "\n",
+        "In this example we use the associated credit card dataset to showcase how you can use AutoML for a simple classification problem. The goal is to predict if a credit card transaction is considered a fraudulent charge.\n",
+        "\n",
+        "This notebook is using local managed compute to train the model.\n",
+        "\n",
+        "If you are using an Azure Machine Learning Compute Instance, you are all set. Otherwise, go through the [configuration](../../../configuration.ipynb) notebook first if you haven't already to establish your connection to the AzureML Workspace. \n",
+        "\n",
+        "In this notebook you will learn how to:\n",
+        "1. Create an experiment using an existing workspace.\n",
+        "2. Configure AutoML using `AutoMLConfig`.\n",
+        "3. Train the model using local managed compute.\n",
+        "4. Explore the results.\n",
+        "5. Test the fitted model."
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "## Setup\n",
+        "\n",
+        "As part of the setup you have already created an Azure ML `Workspace` object. For Automated ML you will need to create an `Experiment` object, which is a named object in a `Workspace` used to run experiments."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "import logging\n",
+        "\n",
+        "import pandas as pd\n",
+        "\n",
+        "import azureml.core\n",
+        "from azureml.core.compute_target import LocalTarget\n",
+        "from azureml.core.experiment import Experiment\n",
+        "from azureml.core.workspace import Workspace\n",
+        "from azureml.core.dataset import Dataset\n",
+        "from azureml.train.automl import AutoMLConfig"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "This sample notebook may use features that are not available in previous versions of the Azure ML SDK."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "print(\"This notebook was created using version 1.38.0 of the Azure ML SDK\")\n",
+        "print(\"You are currently using version\", azureml.core.VERSION, \"of the Azure ML SDK\")"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "ws = Workspace.from_config()\n",
+        "\n",
+        "# choose a name for experiment\n",
+        "experiment_name = 'automl-local-managed'\n",
+        "\n",
+        "experiment=Experiment(ws, experiment_name)\n",
+        "\n",
+        "output = {}\n",
+        "output['Subscription ID'] = ws.subscription_id\n",
+        "output['Workspace'] = ws.name\n",
+        "output['Resource Group'] = ws.resource_group\n",
+        "output['Location'] = ws.location\n",
+        "output['Experiment Name'] = experiment.name\n",
+        "pd.set_option('display.max_colwidth', -1)\n",
+        "outputDf = pd.DataFrame(data = output, index = [''])\n",
+        "outputDf.T"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "### Determine if local docker is configured for Linux images\n",
+        "\n",
+        "Local managed runs will leverage a Linux docker container to submit the run to. Due to this, the docker needs to be configured to use Linux containers."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "# Check if Docker is installed and Linux containers are enabled\n",
+        "import subprocess\n",
+        "from subprocess import CalledProcessError\n",
+        "try:\n",
+        "    assert subprocess.run(\"docker -v\", shell=True).returncode == 0, 'Local Managed runs require docker to be installed.'\n",
+        "    out = subprocess.check_output(\"docker system info\", shell=True).decode('ascii')\n",
+        "    assert \"OSType: linux\" in out, 'Docker engine needs to be configured to use Linux containers.' \\\n",
+        "        'https://docs.docker.com/docker-for-windows/#switch-between-windows-and-linux-containers'\n",
+        "except CalledProcessError as ex:\n",
+        "    raise Exception('Local Managed runs require docker to be installed.') from ex"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "# Data"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "### Load Data\n",
+        "\n",
+        "Load the credit card dataset from a csv file containing both training features and labels. The features are inputs to the model, while the training labels represent the expected output of the model. Next, we'll split the data using random_split and extract the training data for the model."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "data = \"https://automlsamplenotebookdata.blob.core.windows.net/automl-sample-notebook-data/creditcard.csv\"\n",
+        "dataset = Dataset.Tabular.from_delimited_files(data)\n",
+        "training_data, validation_data = dataset.random_split(percentage=0.8, seed=223)\n",
+        "label_column_name = 'Class'"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "## Train\n",
+        "\n",
+        "Instantiate a AutoMLConfig object. This defines the settings and data used to run the experiment.\n",
+        "\n",
+        "|Property|Description|\n",
+        "|-|-|\n",
+        "|**task**|classification or regression|\n",
+        "|**primary_metric**|This is the metric that you want to optimize. Classification supports the following primary metrics: <br><i>accuracy</i><br><i>AUC_weighted</i><br><i>average_precision_score_weighted</i><br><i>norm_macro_recall</i><br><i>precision_score_weighted</i>|\n",
+        "|**enable_early_stopping**|Stop the run if the metric score is not showing improvement.|\n",
+        "|**n_cross_validations**|Number of cross validation splits.|\n",
+        "|**training_data**|Input dataset, containing both features and label column.|\n",
+        "|**label_column_name**|The name of the label column.|\n",
+        "|**enable_local_managed**|Enable the experimental local-managed scenario.|\n",
+        "\n",
+        "**_You can find more information about primary metrics_** [here](https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-configure-auto-train#primary-metric)"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "automl_settings = {\n",
+        "    \"n_cross_validations\": 3,\n",
+        "    \"primary_metric\": 'average_precision_score_weighted',\n",
+        "    \"enable_early_stopping\": True,\n",
+        "    \"experiment_timeout_hours\": 0.3, #for real scenarios we recommend a timeout of at least one hour \n",
+        "    \"verbosity\": logging.INFO,\n",
+        "}\n",
+        "\n",
+        "automl_config = AutoMLConfig(task = 'classification',\n",
+        "                             debug_log = 'automl_errors.log',\n",
+        "                             compute_target = LocalTarget(),\n",
+        "                             enable_local_managed = True,\n",
+        "                             training_data = training_data,\n",
+        "                             label_column_name = label_column_name,\n",
+        "                             **automl_settings\n",
+        "                            )"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "Call the `submit` method on the experiment object and pass the run configuration. Depending on the data and the number of iterations this can run for a while. Validation errors and current status will be shown when setting `show_output=True` and the execution will be synchronous."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "parent_run = experiment.submit(automl_config, show_output = True)"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "# If you need to retrieve a run that already started, use the following code\n",
+        "#from azureml.train.automl.run import AutoMLRun\n",
+        "#parent_run = AutoMLRun(experiment = experiment, run_id = '<replace with your run id>')"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "parent_run"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "## Results"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "#### Explain model\n",
+        "\n",
+        "Automated ML models can be explained and visualized using the SDK Explainability library. "
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "## Analyze results\n",
+        "\n",
+        "### Retrieve the Best Child Run\n",
+        "\n",
+        "Below we select the best pipeline from our iterations. The `get_best_child` method returns the best run. Overloads on `get_best_child` allow you to retrieve the best run for *any* logged metric."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "best_run = parent_run.get_best_child()\n"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "## Test the fitted model\n",
+        "\n",
+        "Now that the model is trained, split the data in the same way the data was split for training (The difference here is the data is being split locally) and then run the test data through the trained model to get the predicted values."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "X_test_df = validation_data.drop_columns(columns=[label_column_name])\n",
+        "y_test_df = validation_data.keep_columns(columns=[label_column_name], validate=True)"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "#### Creating ModelProxy for submitting prediction runs to the training environment.\n",
+        "We will create a ModelProxy for the best child run, which will allow us to submit a run that does the prediction in the training environment. Unlike the local client, which can have different versions of some libraries, the training environment will have all the compatible libraries for the model already."
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "from azureml.train.automl.model_proxy import ModelProxy\n",
+        "best_model_proxy = ModelProxy(best_run)"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "# call the predict functions on the model proxy\n",
+        "y_pred = best_model_proxy.predict(X_test_df).to_pandas_dataframe()\n",
+        "y_pred"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "## Acknowledgements"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "This Credit Card fraud Detection dataset is made available under the Open Database License: http://opendatacommons.org/licenses/odbl/1.0/. Any rights in individual contents of the database are licensed under the Database Contents License: http://opendatacommons.org/licenses/dbcl/1.0/ and is available at: https://www.kaggle.com/mlg-ulb/creditcardfraud\n",
+        "\n",
+        "\n",
+        "The dataset has been collected and analysed during a research collaboration of Worldline and the Machine Learning Group (http://mlg.ulb.ac.be) of ULB (Universit\u00c3\u0192\u00c2\u00a9 Libre de Bruxelles) on big data mining and fraud detection. More details on current and past projects on related topics are available on https://www.researchgate.net/project/Fraud-detection-5 and the page of the DefeatFraud project\n",
+        "Please cite the following works: \n",
+        "\u00c3\u00a2\u00e2\u201a\u00ac\u00c2\u00a2\tAndrea Dal Pozzolo, Olivier Caelen, Reid A. Johnson and Gianluca Bontempi. Calibrating Probability with Undersampling for Unbalanced Classification. In Symposium on Computational Intelligence and Data Mining (CIDM), IEEE, 2015\n",
+        "\u00c3\u00a2\u00e2\u201a\u00ac\u00c2\u00a2\tDal Pozzolo, Andrea; Caelen, Olivier; Le Borgne, Yann-Ael; Waterschoot, Serge; Bontempi, Gianluca. Learned lessons in credit card fraud detection from a practitioner perspective, Expert systems with applications,41,10,4915-4928,2014, Pergamon\n",
+        "\u00c3\u00a2\u00e2\u201a\u00ac\u00c2\u00a2\tDal Pozzolo, Andrea; Boracchi, Giacomo; Caelen, Olivier; Alippi, Cesare; Bontempi, Gianluca. Credit card fraud detection: a realistic modeling and a novel learning strategy, IEEE transactions on neural networks and learning systems,29,8,3784-3797,2018,IEEE\n",
+        "o\tDal Pozzolo, Andrea Adaptive Machine learning for credit card fraud detection ULB MLG PhD thesis (supervised by G. Bontempi)\n",
+        "\u00c3\u00a2\u00e2\u201a\u00ac\u00c2\u00a2\tCarcillo, Fabrizio; Dal Pozzolo, Andrea; Le Borgne, Yann-A\u00c3\u0192\u00c2\u00abl; Caelen, Olivier; Mazzer, Yannis; Bontempi, Gianluca. Scarff: a scalable framework for streaming credit card fraud detection with Spark, Information fusion,41, 182-194,2018,Elsevier\n",
+        "\u00c3\u00a2\u00e2\u201a\u00ac\u00c2\u00a2\tCarcillo, Fabrizio; Le Borgne, Yann-A\u00c3\u0192\u00c2\u00abl; Caelen, Olivier; Bontempi, Gianluca. Streaming active learning strategies for real-life credit card fraud detection: assessment and visualization, International Journal of Data Science and Analytics, 5,4,285-300,2018,Springer International Publishing"
+      ]
+    }
+  ],
+  "metadata": {
+    "authors": [
+      {
+        "name": "sekrupa"
+      }
+    ],
+    "category": "tutorial",
+    "compute": [
+      "AML Compute"
+    ],
+    "datasets": [
+      "Creditcard"
+    ],
+    "deployment": [
+      "None"
+    ],
+    "exclude_from_index": false,
+    "file_extension": ".py",
+    "framework": [
+      "None"
+    ],
+    "friendly_name": "Classification of credit card fraudulent transactions using Automated ML",
+    "index_order": 5,
+    "kernelspec": {
+      "display_name": "Python 3.6",
+      "language": "python",
+      "name": "python36"
+    },
+    "language_info": {
+      "codemirror_mode": {
+        "name": "ipython",
+        "version": 3
+      },
+      "file_extension": ".py",
+      "mimetype": "text/x-python",
+      "name": "python",
+      "nbconvert_exporter": "python",
+      "pygments_lexer": "ipython3",
+      "version": "3.6.7"
+    },
+    "mimetype": "text/x-python",
+    "name": "python",
+    "nbconvert_exporter": "python",
+    "pygments_lexer": "ipython3",
+    "tags": [
+      "AutomatedML"
+    ],
+    "task": "Classification",
+    "version": "3.6.7"
+  },
+  "nbformat": 4,
+  "nbformat_minor": 2
+}

how-to-use-azureml/automated-machine-learning/experimental/classification-credit-card-fraud-local-managed/auto-ml-classification-credit-card-fraud-local-managed.yml

@@ -0,0 +1,4 @@
+name: auto-ml-classification-credit-card-fraud-local-managed
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/experimental/regression-model-proxy/auto-ml-regression-model-proxy.ipynb

@@ -39,6 +39,7 @@
       "source": [
         "## Introduction\n",
         "In this example we use an experimental feature, Model Proxy, to do a predict on the best generated model without downloading the model locally. The prediction will happen on same compute and environment that was used to train the model. This feature is currently in the experimental state, which means that the API is prone to changing, please make sure to run on the latest version of this notebook if you face any issues.\n",
+        "This notebook will also leverage MLFlow for saving models, allowing for more portability of the resulting models. See https://docs.microsoft.com/en-us/azure/machine-learning/how-to-use-mlflow for more details around MLFlow is AzureML.\n",
         "\n",
         "If you are using an Azure Machine Learning Compute Instance, you are all set.  Otherwise, go through the [configuration](../../../../configuration.ipynb)  notebook first if you haven't already to establish your connection to the AzureML Workspace. \n",
         "\n",
@@ -90,7 +91,7 @@
       "metadata": {},
       "outputs": [],
       "source": [
-        "print(\"This notebook was created using version 1.23.0 of the Azure ML SDK\")\n",
+        "print(\"This notebook was created using version 1.38.0 of the Azure ML SDK\")\n",
         "print(\"You are currently using version\", azureml.core.VERSION, \"of the Azure ML SDK\")"
       ]
     },
@@ -142,7 +143,7 @@
         "    compute_target = ComputeTarget(workspace=ws, name=cpu_cluster_name)\n",
         "    print('Found existing cluster, use it.')\n",
         "except ComputeTargetException:\n",
-        "    compute_config = AmlCompute.provisioning_configuration(vm_size='STANDARD_D2_V2',\n",
+        "    compute_config = AmlCompute.provisioning_configuration(vm_size='STANDARD_DS12_V2',\n",
         "                                                           max_nodes=4)\n",
         "    compute_target = ComputeTarget.create(ws, cpu_cluster_name, compute_config)\n",
         "\n",
@@ -194,7 +195,6 @@
         "|**n_cross_validations**|Number of cross validation splits.|\n",
         "|**training_data**|(sparse) array-like, shape = [n_samples, n_features]|\n",
         "|**label_column_name**|(sparse) array-like, shape = [n_samples, ], targets values.|\n",
-        "|**scenario**|We need to set this parameter to 'Latest' to enable some experimental features. This parameter should not be set outside of this experimental notebook.|\n",
         "\n",
         "**_You can find more information about primary metrics_** [here](https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-configure-auto-train#primary-metric)"
       ]
@@ -213,17 +213,17 @@
         "    \"n_cross_validations\": 3,\n",
         "    \"primary_metric\": 'r2_score',\n",
         "    \"enable_early_stopping\": True, \n",
-        "    \"experiment_timeout_hours\": 0.3, #for real scenarios we reccommend a timeout of at least one hour \n",
+        "    \"experiment_timeout_hours\": 0.3, #for real scenarios we recommend a timeout of at least one hour \n",
         "    \"max_concurrent_iterations\": 4,\n",
         "    \"max_cores_per_iteration\": -1,\n",
         "    \"verbosity\": logging.INFO,\n",
+        "    \"save_mlflow\": True,\n",
         "}\n",
         "\n",
         "automl_config = AutoMLConfig(task = 'regression',\n",
         "                             compute_target = compute_target,\n",
         "                             training_data = train_data,\n",
         "                             label_column_name = label,\n",
-        "                             scenario='Latest',\n",
         "                             **automl_settings\n",
         "                            )"
       ]

how-to-use-azureml/automated-machine-learning/experimental/regression-model-proxy/auto-ml-regression-model-proxy.yml

@@ -0,0 +1,4 @@
+name: auto-ml-regression-model-proxy
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/forecasting-backtest-many-models/assets/score.py

@@ -0,0 +1,167 @@
+from typing import Any, Dict, Optional, List
+
+import argparse
+import json
+import os
+import re
+
+import pandas as pd
+
+from matplotlib import pyplot as plt
+from matplotlib.backends.backend_pdf import PdfPages
+
+from azureml.automl.core.shared import constants
+from azureml.automl.core.shared.types import GrainType
+from azureml.automl.runtime.shared.score import scoring
+
+GRAIN = "time_series_id"
+BACKTEST_ITER = "backtest_iteration"
+ACTUALS = "actual_level"
+PREDICTIONS = "predicted_level"
+ALL_GRAINS = "all_sets"
+
+FORECASTS_FILE = "forecast.csv"
+SCORES_FILE = "scores.csv"
+PLOTS_FILE = "plots_fcst_vs_actual.pdf"
+RE_INVALID_SYMBOLS = re.compile("[: ]")
+
+
+def _compute_metrics(df: pd.DataFrame, metrics: List[str]):
+    """
+    Compute metrics for one data frame.
+
+    :param df: The data frame which contains actual_level and predicted_level columns.
+    :return: The data frame with two columns - metric_name and metric.
+    """
+    scores = scoring.score_regression(
+        y_test=df[ACTUALS], y_pred=df[PREDICTIONS], metrics=metrics
+    )
+    metrics_df = pd.DataFrame(list(scores.items()), columns=["metric_name", "metric"])
+    metrics_df.sort_values(["metric_name"], inplace=True)
+    metrics_df.reset_index(drop=True, inplace=True)
+    return metrics_df
+
+
+def _format_grain_name(grain: GrainType) -> str:
+    """
+    Convert grain name to string.
+
+    :param grain: the grain name.
+    :return: the string representation of the given grain.
+    """
+    if not isinstance(grain, tuple) and not isinstance(grain, list):
+        return str(grain)
+    grain = list(map(str, grain))
+    return "|".join(grain)
+
+
+def compute_all_metrics(
+    fcst_df: pd.DataFrame,
+    ts_id_colnames: List[str],
+    metric_names: Optional[List[set]] = None,
+):
+    """
+    Calculate metrics per grain.
+
+    :param fcst_df: forecast data frame. Must contain 2 columns: 'actual_level' and 'predicted_level'
+    :param metric_names: (optional) the list of metric names to return
+    :param ts_id_colnames: (optional) list of grain column names
+    :return: dictionary of summary table for all tests and final decision on stationary vs nonstaionary
+    """
+    if not metric_names:
+        metric_names = list(constants.Metric.SCALAR_REGRESSION_SET)
+
+    if ts_id_colnames is None:
+        ts_id_colnames = []
+
+    metrics_list = []
+    if ts_id_colnames:
+        for grain, df in fcst_df.groupby(ts_id_colnames):
+            one_grain_metrics_df = _compute_metrics(df, metric_names)
+            one_grain_metrics_df[GRAIN] = _format_grain_name(grain)
+            metrics_list.append(one_grain_metrics_df)
+
+    # overall metrics
+    one_grain_metrics_df = _compute_metrics(fcst_df, metric_names)
+    one_grain_metrics_df[GRAIN] = ALL_GRAINS
+    metrics_list.append(one_grain_metrics_df)
+
+    # collect into a data frame
+    return pd.concat(metrics_list)
+
+
+def _draw_one_plot(
+    df: pd.DataFrame,
+    time_column_name: str,
+    grain_column_names: List[str],
+    pdf: PdfPages,
+) -> None:
+    """
+    Draw the single plot.
+
+    :param df: The data frame with the data to build plot.
+    :param time_column_name: The name of a time column.
+    :param grain_column_names: The name of grain columns.
+    :param pdf: The pdf backend used to render the plot.
+    """
+    fig, _ = plt.subplots(figsize=(20, 10))
+    df = df.set_index(time_column_name)
+    plt.plot(df[[ACTUALS, PREDICTIONS]])
+    plt.xticks(rotation=45)
+    iteration = df[BACKTEST_ITER].iloc[0]
+    if grain_column_names:
+        grain_name = [df[grain].iloc[0] for grain in grain_column_names]
+        plt.title(f"Time series ID: {_format_grain_name(grain_name)} {iteration}")
+    plt.legend(["actual", "forecast"])
+    plt.close(fig)
+    pdf.savefig(fig)
+
+
+def calculate_scores_and_build_plots(
+    input_dir: str, output_dir: str, automl_settings: Dict[str, Any]
+):
+    os.makedirs(output_dir, exist_ok=True)
+    grains = automl_settings.get(constants.TimeSeries.GRAIN_COLUMN_NAMES)
+    time_column_name = automl_settings.get(constants.TimeSeries.TIME_COLUMN_NAME)
+    if grains is None:
+        grains = []
+    if isinstance(grains, str):
+        grains = [grains]
+    while BACKTEST_ITER in grains:
+        grains.remove(BACKTEST_ITER)
+
+    dfs = []
+    for fle in os.listdir(input_dir):
+        file_path = os.path.join(input_dir, fle)
+        if os.path.isfile(file_path) and file_path.endswith(".csv"):
+            df_iter = pd.read_csv(file_path, parse_dates=[time_column_name])
+            for _, iteration in df_iter.groupby(BACKTEST_ITER):
+                dfs.append(iteration)
+    forecast_df = pd.concat(dfs, sort=False, ignore_index=True)
+    # To make sure plots are in order, sort the predictions by grain and iteration.
+    ts_index = grains + [BACKTEST_ITER]
+    forecast_df.sort_values(by=ts_index, inplace=True)
+    pdf = PdfPages(os.path.join(output_dir, PLOTS_FILE))
+    for _, one_forecast in forecast_df.groupby(ts_index):
+        _draw_one_plot(one_forecast, time_column_name, grains, pdf)
+    pdf.close()
+    forecast_df.to_csv(os.path.join(output_dir, FORECASTS_FILE), index=False)
+    metrics = compute_all_metrics(forecast_df, grains + [BACKTEST_ITER])
+    metrics.to_csv(os.path.join(output_dir, SCORES_FILE), index=False)
+
+
+if __name__ == "__main__":
+    args = {"forecasts": "--forecasts", "scores_out": "--output-dir"}
+    parser = argparse.ArgumentParser("Parsing input arguments.")
+    for argname, arg in args.items():
+        parser.add_argument(arg, dest=argname, required=True)
+    parsed_args, _ = parser.parse_known_args()
+    input_dir = parsed_args.forecasts
+    output_dir = parsed_args.scores_out
+    with open(
+        os.path.join(
+            os.path.dirname(os.path.realpath(__file__)), "automl_settings.json"
+        )
+    ) as json_file:
+        automl_settings = json.load(json_file)
+    calculate_scores_and_build_plots(input_dir, output_dir, automl_settings)

how-to-use-azureml/automated-machine-learning/forecasting-backtest-many-models/auto-ml-forecasting-backtest-many-models.ipynb

@@ -0,0 +1,725 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Copyright (c) Microsoft Corporation. All rights reserved.\n",
+    "\n",
+    "Licensed under the MIT License."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "![Impressions](https://PixelServer20190423114238.azurewebsites.net/api/impressions/MachineLearningNotebooks/how-to-use-azureml/automated-machine-learning/forecasting-hierarchical-timeseries/auto-ml-forecasting-hierarchical-timeseries.png)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Many Models with Backtesting - Automated ML\n",
+    "**_Backtest many models time series forecasts with Automated Machine Learning_**\n",
+    "\n",
+    "---"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "For this notebook we are using a synthetic dataset to demonstrate the back testing in many model scenario. This allows us to check historical performance of AutoML on a historical data. To do that we step back on the backtesting period by the data set several times and split the data to train and test sets. Then these data sets are used for training and evaluation of model.<br>\n",
+    "\n",
+    "Thus, it is a quick way of evaluating AutoML as if it was in production. Here, we do not test historical performance of a particular model, for this see the [notebook](../forecasting-backtest-single-model/auto-ml-forecasting-backtest-single-model.ipynb). Instead, the best model for every backtest iteration can be different since AutoML chooses the best model for a given training set.\n",
+    "![Backtesting](Backtesting.png)\n",
+    "\n",
+    "**NOTE: There are limits on how many runs we can do in parallel per workspace, and we currently recommend to set the parallelism to maximum of 320 runs per experiment per workspace. If users want to have more parallelism and increase this limit they might encounter Too Many Requests errors (HTTP 429).**"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Prerequisites\n",
+    "You'll need to create a compute Instance by following the instructions in the [EnvironmentSetup.md](../Setup_Resources/EnvironmentSetup.md)."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 1.0 Set up workspace, datastore, experiment"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613003526897
+    }
+   },
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "\n",
+    "import azureml.core\n",
+    "from azureml.core import Workspace, Datastore\n",
+    "import numpy as np\n",
+    "import pandas as pd\n",
+    "\n",
+    "from pandas.tseries.frequencies import to_offset\n",
+    "\n",
+    "# Set up your workspace\n",
+    "ws = Workspace.from_config()\n",
+    "ws.get_details()\n",
+    "\n",
+    "# Set up your datastores\n",
+    "dstore = ws.get_default_datastore()\n",
+    "\n",
+    "output = {}\n",
+    "output[\"SDK version\"] = azureml.core.VERSION\n",
+    "output[\"Subscription ID\"] = ws.subscription_id\n",
+    "output[\"Workspace\"] = ws.name\n",
+    "output[\"Resource Group\"] = ws.resource_group\n",
+    "output[\"Location\"] = ws.location\n",
+    "output[\"Default datastore name\"] = dstore.name\n",
+    "pd.set_option(\"display.max_colwidth\", -1)\n",
+    "outputDf = pd.DataFrame(data=output, index=[\"\"])\n",
+    "outputDf.T"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "This notebook is compatible with Azure ML SDK version 1.35.1 or later."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "print(\"You are currently using version\", azureml.core.VERSION, \"of the Azure ML SDK\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Choose an experiment"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613003540729
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.core import Experiment\n",
+    "\n",
+    "experiment = Experiment(ws, \"automl-many-models-backtest\")\n",
+    "\n",
+    "print(\"Experiment name: \" + experiment.name)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 2.0 Data\n",
+    "\n",
+    "#### 2.1 Data generation\n",
+    "For this notebook we will generate the artificial data set with two [time series IDs](https://docs.microsoft.com/en-us/python/api/azureml-automl-core/azureml.automl.core.forecasting_parameters.forecastingparameters?view=azure-ml-py). Then we will generate backtest folds and will upload it to the default BLOB storage and create a [TabularDataset](https://docs.microsoft.com/en-us/python/api/azureml-core/azureml.data.tabular_dataset.tabulardataset?view=azure-ml-py)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# simulate data: 2 grains - 700\n",
+    "TIME_COLNAME = \"date\"\n",
+    "TARGET_COLNAME = \"value\"\n",
+    "TIME_SERIES_ID_COLNAME = \"ts_id\"\n",
+    "\n",
+    "sample_size = 700\n",
+    "# Set the random seed for reproducibility of results.\n",
+    "np.random.seed(20)\n",
+    "X1 = pd.DataFrame(\n",
+    "    {\n",
+    "        TIME_COLNAME: pd.date_range(start=\"2018-01-01\", periods=sample_size),\n",
+    "        TARGET_COLNAME: np.random.normal(loc=100, scale=20, size=sample_size),\n",
+    "        TIME_SERIES_ID_COLNAME: \"ts_A\",\n",
+    "    }\n",
+    ")\n",
+    "X2 = pd.DataFrame(\n",
+    "    {\n",
+    "        TIME_COLNAME: pd.date_range(start=\"2018-01-01\", periods=sample_size),\n",
+    "        TARGET_COLNAME: np.random.normal(loc=100, scale=20, size=sample_size),\n",
+    "        TIME_SERIES_ID_COLNAME: \"ts_B\",\n",
+    "    }\n",
+    ")\n",
+    "\n",
+    "X = pd.concat([X1, X2], ignore_index=True, sort=False)\n",
+    "print(\"Simulated dataset contains {} rows \\n\".format(X.shape[0]))\n",
+    "X.head()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Now we will generate 8 backtesting folds with backtesting period of 7 days and with the same forecasting horizon. We will add the column \"backtest_iteration\", which will identify the backtesting period by the last training date."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "offset_type = \"7D\"\n",
+    "NUMBER_OF_BACKTESTS = 8  # number of train/test sets to generate\n",
+    "\n",
+    "dfs_train = []\n",
+    "dfs_test = []\n",
+    "for ts_id, df_one in X.groupby(TIME_SERIES_ID_COLNAME):\n",
+    "\n",
+    "    data_end = df_one[TIME_COLNAME].max()\n",
+    "\n",
+    "    for i in range(NUMBER_OF_BACKTESTS):\n",
+    "        train_cutoff_date = data_end - to_offset(offset_type)\n",
+    "        df_one = df_one.copy()\n",
+    "        df_one[\"backtest_iteration\"] = \"iteration_\" + str(train_cutoff_date)\n",
+    "        train = df_one[df_one[TIME_COLNAME] <= train_cutoff_date]\n",
+    "        test = df_one[\n",
+    "            (df_one[TIME_COLNAME] > train_cutoff_date)\n",
+    "            & (df_one[TIME_COLNAME] <= data_end)\n",
+    "        ]\n",
+    "        data_end = train[TIME_COLNAME].max()\n",
+    "        dfs_train.append(train)\n",
+    "        dfs_test.append(test)\n",
+    "\n",
+    "X_train = pd.concat(dfs_train, sort=False, ignore_index=True)\n",
+    "X_test = pd.concat(dfs_test, sort=False, ignore_index=True)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### 2.2 Create the Tabular Data Set.\n",
+    "\n",
+    "A Datastore is a place where data can be stored that is then made accessible to a compute either by means of mounting or copying the data to the compute target.\n",
+    "\n",
+    "Please refer to [Datastore](https://docs.microsoft.com/en-us/python/api/azureml-core/azureml.core.datastore(class)?view=azure-ml-py) documentation on how to access data from Datastore.\n",
+    "\n",
+    "In this next step, we will upload the data and create a TabularDataset."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.data.dataset_factory import TabularDatasetFactory\n",
+    "\n",
+    "ds = ws.get_default_datastore()\n",
+    "# Upload saved data to the default data store.\n",
+    "train_data = TabularDatasetFactory.register_pandas_dataframe(\n",
+    "    X_train, target=(ds, \"data_mm\"), name=\"data_train\"\n",
+    ")\n",
+    "test_data = TabularDatasetFactory.register_pandas_dataframe(\n",
+    "    X_test, target=(ds, \"data_mm\"), name=\"data_test\"\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 3.0 Build the training pipeline\n",
+    "Now that the dataset, WorkSpace, and datastore are set up, we can put together a pipeline for training.\n",
+    "\n",
+    "> Note that if you have an AzureML Data Scientist role, you will not have permission to create compute resources. Talk to your workspace or IT admin to create the compute targets described in this section, if they do not already exist."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Choose a compute target\n",
+    "\n",
+    "You will need to create a [compute target](https://docs.microsoft.com/en-us/azure/machine-learning/how-to-set-up-training-targets#amlcompute) for your AutoML run. In this tutorial, you create AmlCompute as your training compute resource.\n",
+    "\n",
+    "\\*\\*Creation of AmlCompute takes approximately 5 minutes.**\n",
+    "\n",
+    "If the AmlCompute with that name is already in your workspace this code will skip the creation process. As with other Azure services, there are limits on certain resources (e.g. AmlCompute) associated with the Azure Machine Learning service. Please read this [article](https://docs.microsoft.com/en-us/azure/machine-learning/how-to-manage-quotas) on the default limits and how to request more quota."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613007037308
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.core.compute import ComputeTarget, AmlCompute\n",
+    "\n",
+    "# Name your cluster\n",
+    "compute_name = \"backtest-mm\"\n",
+    "\n",
+    "\n",
+    "if compute_name in ws.compute_targets:\n",
+    "    compute_target = ws.compute_targets[compute_name]\n",
+    "    if compute_target and type(compute_target) is AmlCompute:\n",
+    "        print(\"Found compute target: \" + compute_name)\n",
+    "else:\n",
+    "    print(\"Creating a new compute target...\")\n",
+    "    provisioning_config = AmlCompute.provisioning_configuration(\n",
+    "        vm_size=\"STANDARD_DS12_V2\", max_nodes=6\n",
+    "    )\n",
+    "    # Create the compute target\n",
+    "    compute_target = ComputeTarget.create(ws, compute_name, provisioning_config)\n",
+    "\n",
+    "    # Can poll for a minimum number of nodes and for a specific timeout.\n",
+    "    # If no min node count is provided it will use the scale settings for the cluster\n",
+    "    compute_target.wait_for_completion(\n",
+    "        show_output=True, min_node_count=None, timeout_in_minutes=20\n",
+    "    )\n",
+    "\n",
+    "    # For a more detailed view of current cluster status, use the 'status' property\n",
+    "    print(compute_target.status.serialize())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Set up training parameters\n",
+    "\n",
+    "This dictionary defines the AutoML and many models settings. For this forecasting task we need to define several settings including the name of the time column, the maximum forecast horizon, and the partition column name definition. Please note, that in this case we are setting grain_column_names to be the time series ID column plus iteration, because we want to train a separate model for each time series and iteration.\n",
+    "\n",
+    "| Property                           | Description|\n",
+    "| :---------------                   | :------------------- |\n",
+    "| **task**                           | forecasting |\n",
+    "| **primary_metric**                 | This is the metric that you want to optimize.<br> Forecasting supports the following primary metrics <br><i>normalized_root_mean_squared_error</i><br><i>normalized_mean_absolute_error</i> |\n",
+    "| **iteration_timeout_minutes**      | Maximum amount of time in minutes that the model can train. This is optional but provides customers with greater control on exit criteria. |\n",
+    "| **iterations**                     | Number of models to train. This is optional but provides customers with greater control on exit criteria. |\n",
+    "| **experiment_timeout_hours**       | Maximum amount of time in hours that the experiment can take before it terminates. This is optional but provides customers with greater control on exit criteria. |\n",
+    "| **label_column_name**              | The name of the label column. |\n",
+    "| **max_horizon**               | The forecast horizon is how many periods forward you would like to forecast. This integer horizon is in units of the timeseries frequency (e.g. daily, weekly). Periods are inferred from your data. |\n",
+    "| **n_cross_validations**            | Number of cross validation splits. Rolling Origin Validation is used to split time-series in a temporally consistent way. |\n",
+    "| **time_column_name**               | The name of your time column. |\n",
+    "| **grain_column_names**     | The column names used to uniquely identify timeseries in data that has multiple rows with the same timestamp. |\n",
+    "| **track_child_runs**               | Flag to disable tracking of child runs. Only best run is tracked if the flag is set to False (this includes the model and metrics of the run). |\n",
+    "| **partition_column_names**         | The names of columns used to group your models. For timeseries, the groups must not split up individual time-series. That is, each group must contain one or more whole time-series. |"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613007061544
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.train.automl.runtime._many_models.many_models_parameters import (\n",
+    "    ManyModelsTrainParameters,\n",
+    ")\n",
+    "\n",
+    "partition_column_names = [TIME_SERIES_ID_COLNAME, \"backtest_iteration\"]\n",
+    "automl_settings = {\n",
+    "    \"task\": \"forecasting\",\n",
+    "    \"primary_metric\": \"normalized_root_mean_squared_error\",\n",
+    "    \"iteration_timeout_minutes\": 10,  # This needs to be changed based on the dataset. We ask customer to explore how long training is taking before settings this value\n",
+    "    \"iterations\": 15,\n",
+    "    \"experiment_timeout_hours\": 0.25,  # This also needs to be changed based on the dataset. For larger data set this number needs to be bigger.\n",
+    "    \"label_column_name\": TARGET_COLNAME,\n",
+    "    \"n_cross_validations\": 3,\n",
+    "    \"time_column_name\": TIME_COLNAME,\n",
+    "    \"max_horizon\": 6,\n",
+    "    \"grain_column_names\": partition_column_names,\n",
+    "    \"track_child_runs\": False,\n",
+    "}\n",
+    "\n",
+    "mm_paramters = ManyModelsTrainParameters(\n",
+    "    automl_settings=automl_settings, partition_column_names=partition_column_names\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Set up many models pipeline"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Parallel run step is leveraged to train multiple models at once. To configure the ParallelRunConfig you will need to determine the appropriate number of workers and nodes for your use case. The process_count_per_node is based off the number of cores of the compute VM. The node_count will determine the number of master nodes to use, increasing the node count will speed up the training process.\n",
+    "\n",
+    "| Property                           | Description|\n",
+    "| :---------------                   | :------------------- |\n",
+    "| **experiment**                     | The experiment used for training. |\n",
+    "| **train_data**                     | The file dataset to be used as input to the training run. |\n",
+    "| **node_count**                     | The number of compute nodes to be used for running the user script. We recommend to start with 3 and increase the node_count if the training time is taking too long. |\n",
+    "| **process_count_per_node**         | Process count per node, we recommend 2:1 ratio for number of cores: number of processes per node. eg. If node has 16 cores then configure 8 or less process count per node or optimal performance. |\n",
+    "| **train_pipeline_parameters**      | The set of configuration parameters defined in the previous section. |\n",
+    "\n",
+    "Calling this method will create a new aggregated dataset which is generated dynamically on pipeline execution."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.contrib.automl.pipeline.steps import AutoMLPipelineBuilder\n",
+    "\n",
+    "\n",
+    "training_pipeline_steps = AutoMLPipelineBuilder.get_many_models_train_steps(\n",
+    "    experiment=experiment,\n",
+    "    train_data=train_data,\n",
+    "    compute_target=compute_target,\n",
+    "    node_count=2,\n",
+    "    process_count_per_node=2,\n",
+    "    run_invocation_timeout=920,\n",
+    "    train_pipeline_parameters=mm_paramters,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.pipeline.core import Pipeline\n",
+    "\n",
+    "training_pipeline = Pipeline(ws, steps=training_pipeline_steps)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Submit the pipeline to run\n",
+    "Next we submit our pipeline to run. The whole training pipeline takes about 20 minutes using a STANDARD_DS12_V2 VM with our current ParallelRunConfig setting."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "training_run = experiment.submit(training_pipeline)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "training_run.wait_for_completion(show_output=False)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Check the run status, if training_run is in completed state, continue to next section. Otherwise, check the portal for failures."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 4.0 Backtesting\n",
+    "Now that we selected the best AutoML model for each backtest fold, we will use these models to generate the forecasts and compare with the actuals."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Set up output dataset for inference data\n",
+    "Output of inference can be represented as [OutputFileDatasetConfig](https://docs.microsoft.com/en-us/python/api/azureml-core/azureml.data.output_dataset_config.outputdatasetconfig?view=azure-ml-py) object and OutputFileDatasetConfig can be registered as a dataset. "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.data import OutputFileDatasetConfig\n",
+    "\n",
+    "output_inference_data_ds = OutputFileDatasetConfig(\n",
+    "    name=\"many_models_inference_output\",\n",
+    "    destination=(dstore, \"backtesting/inference_data/\"),\n",
+    ").register_on_complete(name=\"backtesting_data_ds\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "For many models we need to provide the ManyModelsInferenceParameters object.\n",
+    "\n",
+    "#### ManyModelsInferenceParameters arguments\n",
+    "| Property                           | Description|\n",
+    "| :---------------                   | :------------------- |\n",
+    "| **partition_column_names**         | List of column names that identifies groups.                                           |\n",
+    "| **target_column_name**             | \\[Optional\\] Column name only if the inference dataset has the target. |\n",
+    "| **time_column_name**               | Column name only if it is timeseries. |\n",
+    "| **many_models_run_id**             | \\[Optional\\] Many models pipeline run id where models were trained. |\n",
+    "\n",
+    "#### get_many_models_batch_inference_steps arguments\n",
+    "| Property                           | Description|\n",
+    "| :---------------                   | :------------------- |\n",
+    "| **experiment**                     | The experiment used for inference run. |\n",
+    "| **inference_data**                 | The data to use for inferencing. It should be the same schema as used for training.\n",
+    "| **compute_target**                 | The compute target that runs the inference pipeline.|\n",
+    "| **node_count**                     | The number of compute nodes to be used for running the user script. We recommend to start with the number of cores per node (varies by compute sku). |\n",
+    "| **process_count_per_node**         | The number of processes per node.\n",
+    "| **train_run_id**                   | \\[Optional\\] The run id of the hierarchy training, by default it is the latest successful training many model run in the experiment. |\n",
+    "| **train_experiment_name**          | \\[Optional\\] The train experiment that contains the train pipeline. This one is only needed when the train pipeline is not in the same experiement as the inference pipeline. |\n",
+    "| **process_count_per_node**         | \\[Optional\\] The number of processes per node, by default it's 4. |"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.contrib.automl.pipeline.steps import AutoMLPipelineBuilder\n",
+    "from azureml.train.automl.runtime._many_models.many_models_parameters import (\n",
+    "    ManyModelsInferenceParameters,\n",
+    ")\n",
+    "\n",
+    "mm_parameters = ManyModelsInferenceParameters(\n",
+    "    partition_column_names=partition_column_names,\n",
+    "    time_column_name=TIME_COLNAME,\n",
+    "    target_column_name=TARGET_COLNAME,\n",
+    ")\n",
+    "\n",
+    "inference_steps = AutoMLPipelineBuilder.get_many_models_batch_inference_steps(\n",
+    "    experiment=experiment,\n",
+    "    inference_data=test_data,\n",
+    "    node_count=2,\n",
+    "    process_count_per_node=2,\n",
+    "    compute_target=compute_target,\n",
+    "    run_invocation_timeout=300,\n",
+    "    output_datastore=output_inference_data_ds,\n",
+    "    train_run_id=training_run.id,\n",
+    "    train_experiment_name=training_run.experiment.name,\n",
+    "    inference_pipeline_parameters=mm_parameters,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.pipeline.core import Pipeline\n",
+    "\n",
+    "inference_pipeline = Pipeline(ws, steps=inference_steps)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "inference_run = experiment.submit(inference_pipeline)\n",
+    "inference_run.wait_for_completion(show_output=False)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 5.0 Retrieve results and calculate metrics\n",
+    "\n",
+    "The pipeline returns one file with the predictions for each times series ID and outputs the result to the forecasting_output Blob container. The details of the blob container is listed in 'forecasting_output.txt' under Outputs+logs. \n",
+    "\n",
+    "The next code snippet does the following:\n",
+    "1. Downloads the contents of the output folder that is passed in the parallel run step \n",
+    "2. Reads the parallel_run_step.txt file that has the predictions as pandas dataframe \n",
+    "3. Saves the table in csv format and \n",
+    "4. Displays the top 10 rows of the predictions"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.contrib.automl.pipeline.steps.utilities import get_output_from_mm_pipeline\n",
+    "\n",
+    "forecasting_results_name = \"forecasting_results\"\n",
+    "forecasting_output_name = \"many_models_inference_output\"\n",
+    "forecast_file = get_output_from_mm_pipeline(\n",
+    "    inference_run, forecasting_results_name, forecasting_output_name\n",
+    ")\n",
+    "df = pd.read_csv(forecast_file, delimiter=\" \", header=None, parse_dates=[0])\n",
+    "df.columns = list(X_train.columns) + [\"predicted_level\"]\n",
+    "print(\n",
+    "    \"Prediction has \", df.shape[0], \" rows. Here the first 10 rows are being displayed.\"\n",
+    ")\n",
+    "# Save the scv file with header to read it in the next step.\n",
+    "df.rename(columns={TARGET_COLNAME: \"actual_level\"}, inplace=True)\n",
+    "df.to_csv(os.path.join(forecasting_results_name, \"forecast.csv\"), index=False)\n",
+    "df.head(10)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## View metrics\n",
+    "We will read in the obtained results and run the helper script, which will generate metrics and create the plots of predicted versus actual values."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from assets.score import calculate_scores_and_build_plots\n",
+    "\n",
+    "backtesting_results = \"backtesting_mm_results\"\n",
+    "os.makedirs(backtesting_results, exist_ok=True)\n",
+    "calculate_scores_and_build_plots(\n",
+    "    forecasting_results_name, backtesting_results, automl_settings\n",
+    ")\n",
+    "pd.DataFrame({\"File\": os.listdir(backtesting_results)})"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "The directory contains a set of files with results:\n",
+    "- forecast.csv contains forecasts for all backtest iterations. The backtest_iteration column contains iteration identifier with the last training date as a suffix\n",
+    "- scores.csv contains all metrics. If data set contains several time series, the metrics are given for all combinations of time series id and iterations, as well as scores for all iterations and time series ids, which are marked as \"all_sets\"\n",
+    "- plots_fcst_vs_actual.pdf contains the predictions vs forecast plots for each iteration and, eash time series is saved as separate plot.\n",
+    "\n",
+    "For demonstration purposes we will display the table of metrics for one of the time series with ID \"ts0\". We will create the utility function, which will build the table with metrics."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def get_metrics_for_ts(all_metrics, ts):\n",
+    "    \"\"\"\n",
+    "    Get the metrics for the time series with ID ts and return it as pandas data frame.\n",
+    "\n",
+    "    :param all_metrics: The table with all the metrics.\n",
+    "    :param ts: The ID of a time series of interest.\n",
+    "    :return: The pandas DataFrame with metrics for one time series.\n",
+    "    \"\"\"\n",
+    "    results_df = None\n",
+    "    for ts_id, one_series in all_metrics.groupby(\"time_series_id\"):\n",
+    "        if not ts_id.startswith(ts):\n",
+    "            continue\n",
+    "        iteration = ts_id.split(\"|\")[-1]\n",
+    "        df = one_series[[\"metric_name\", \"metric\"]]\n",
+    "        df.rename({\"metric\": iteration}, axis=1, inplace=True)\n",
+    "        df.set_index(\"metric_name\", inplace=True)\n",
+    "        if results_df is None:\n",
+    "            results_df = df\n",
+    "        else:\n",
+    "            results_df = results_df.merge(\n",
+    "                df, how=\"inner\", left_index=True, right_index=True\n",
+    "            )\n",
+    "    results_df.sort_index(axis=1, inplace=True)\n",
+    "    return results_df\n",
+    "\n",
+    "\n",
+    "metrics_df = pd.read_csv(os.path.join(backtesting_results, \"scores.csv\"))\n",
+    "ts = \"ts_A\"\n",
+    "get_metrics_for_ts(metrics_df, ts)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Forecast vs actuals plots."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from IPython.display import IFrame\n",
+    "\n",
+    "IFrame(\"./backtesting_mm_results/plots_fcst_vs_actual.pdf\", width=800, height=300)"
+   ]
+  }
+ ],
+ "metadata": {
+  "authors": [
+   {
+    "name": "jialiu"
+   }
+  ],
+  "categories": [
+   "how-to-use-azureml",
+   "automated-machine-learning"
+  ],
+  "kernelspec": {
+   "display_name": "Python 3.6 - AzureML",
+   "language": "python",
+   "name": "python3-azureml"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.6.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}

how-to-use-azureml/automated-machine-learning/forecasting-backtest-many-models/auto-ml-forecasting-backtest-many-models.yml

@@ -0,0 +1,4 @@
+name: auto-ml-forecasting-backtest-many-models
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/forecasting-backtest-many-models/update_env.yml

@@ -0,0 +1,3 @@
+dependencies:
+- pip:
+  - azureml-contrib-automl-pipeline-steps

how-to-use-azureml/automated-machine-learning/forecasting-backtest-single-model/assets/data_split.py

@@ -0,0 +1,45 @@
+import argparse
+import os
+
+import pandas as pd
+
+import azureml.train.automl.runtime._hts.hts_runtime_utilities as hru
+
+from azureml.core import Run
+from azureml.core.dataset import Dataset
+
+# Parse the arguments.
+args = {
+    "step_size": "--step-size",
+    "step_number": "--step-number",
+    "time_column_name": "--time-column-name",
+    "time_series_id_column_names": "--time-series-id-column-names",
+    "out_dir": "--output-dir",
+}
+parser = argparse.ArgumentParser("Parsing input arguments.")
+for argname, arg in args.items():
+    parser.add_argument(arg, dest=argname, required=True)
+parsed_args, _ = parser.parse_known_args()
+step_number = int(parsed_args.step_number)
+step_size = int(parsed_args.step_size)
+# Create the working dirrectory to store the temporary csv files.
+working_dir = parsed_args.out_dir
+os.makedirs(working_dir, exist_ok=True)
+# Set input and output
+script_run = Run.get_context()
+input_dataset = script_run.input_datasets["training_data"]
+X_train = input_dataset.to_pandas_dataframe()
+# Split the data.
+for i in range(step_number):
+    file_name = os.path.join(working_dir, "backtest_{}.csv".format(i))
+    if parsed_args.time_series_id_column_names:
+        dfs = []
+        for _, one_series in X_train.groupby([parsed_args.time_series_id_column_names]):
+            one_series = one_series.sort_values(
+                by=[parsed_args.time_column_name], inplace=False
+            )
+            dfs.append(one_series.iloc[: len(one_series) - step_size * i])
+        pd.concat(dfs, sort=False, ignore_index=True).to_csv(file_name, index=False)
+    else:
+        X_train.sort_values(by=[parsed_args.time_column_name], inplace=True)
+        X_train.iloc[: len(X_train) - step_size * i].to_csv(file_name, index=False)

how-to-use-azureml/automated-machine-learning/forecasting-backtest-single-model/assets/retrain_models.py

@@ -0,0 +1,173 @@
+# ---------------------------------------------------------
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# ---------------------------------------------------------
+"""The batch script needed for back testing of models using PRS."""
+import argparse
+import json
+import logging
+import os
+import pickle
+import re
+
+import pandas as pd
+
+from azureml.core.experiment import Experiment
+from azureml.core.model import Model
+from azureml.core.run import Run
+from azureml.automl.core.shared import constants
+from azureml.automl.runtime.shared.score import scoring
+from azureml.train.automl import AutoMLConfig
+
+RE_INVALID_SYMBOLS = re.compile(r"[:\s]")
+
+model_name = None
+target_column_name = None
+current_step_run = None
+output_dir = None
+
+logger = logging.getLogger(__name__)
+
+
+def _get_automl_settings():
+    with open(
+        os.path.join(
+            os.path.dirname(os.path.realpath(__file__)), "automl_settings.json"
+        )
+    ) as json_file:
+        return json.load(json_file)
+
+
+def init():
+    global model_name
+    global target_column_name
+    global output_dir
+    global automl_settings
+    global model_uid
+    logger.info("Initialization of the run.")
+    parser = argparse.ArgumentParser("Parsing input arguments.")
+    parser.add_argument("--output-dir", dest="out", required=True)
+    parser.add_argument("--model-name", dest="model", default=None)
+    parser.add_argument("--model-uid", dest="model_uid", default=None)
+
+    parsed_args, _ = parser.parse_known_args()
+    model_name = parsed_args.model
+    automl_settings = _get_automl_settings()
+    target_column_name = automl_settings.get("label_column_name")
+    output_dir = parsed_args.out
+    model_uid = parsed_args.model_uid
+    os.makedirs(output_dir, exist_ok=True)
+    os.environ["AUTOML_IGNORE_PACKAGE_VERSION_INCOMPATIBILITIES".lower()] = "True"
+
+
+def get_run():
+    global current_step_run
+    if current_step_run is None:
+        current_step_run = Run.get_context()
+    return current_step_run
+
+
+def run_backtest(data_input_name: str, file_name: str, experiment: Experiment):
+    """Re-train the model and return metrics."""
+    data_input = pd.read_csv(
+        data_input_name,
+        parse_dates=[automl_settings[constants.TimeSeries.TIME_COLUMN_NAME]],
+    )
+    print(data_input.head())
+    if not automl_settings.get(constants.TimeSeries.GRAIN_COLUMN_NAMES):
+        # There is no grains.
+        data_input.sort_values(
+            [automl_settings[constants.TimeSeries.TIME_COLUMN_NAME]], inplace=True
+        )
+        X_train = data_input.iloc[: -automl_settings["max_horizon"]]
+        y_train = X_train.pop(target_column_name).values
+        X_test = data_input.iloc[-automl_settings["max_horizon"] :]
+        y_test = X_test.pop(target_column_name).values
+    else:
+        # The data contain grains.
+        dfs_train = []
+        dfs_test = []
+        for _, one_series in data_input.groupby(
+            automl_settings.get(constants.TimeSeries.GRAIN_COLUMN_NAMES)
+        ):
+            one_series.sort_values(
+                [automl_settings[constants.TimeSeries.TIME_COLUMN_NAME]], inplace=True
+            )
+            dfs_train.append(one_series.iloc[: -automl_settings["max_horizon"]])
+            dfs_test.append(one_series.iloc[-automl_settings["max_horizon"] :])
+        X_train = pd.concat(dfs_train, sort=False, ignore_index=True)
+        y_train = X_train.pop(target_column_name).values
+        X_test = pd.concat(dfs_test, sort=False, ignore_index=True)
+        y_test = X_test.pop(target_column_name).values
+
+    last_training_date = str(
+        X_train[automl_settings[constants.TimeSeries.TIME_COLUMN_NAME]].max()
+    )
+
+    if file_name:
+        # If file name is provided, we will load model and retrain it on backtest data.
+        with open(file_name, "rb") as fp:
+            fitted_model = pickle.load(fp)
+        fitted_model.fit(X_train, y_train)
+    else:
+        # We will run the experiment and select the best model.
+        X_train[target_column_name] = y_train
+        automl_config = AutoMLConfig(training_data=X_train, **automl_settings)
+        automl_run = current_step_run.submit_child(automl_config, show_output=True)
+        best_run, fitted_model = automl_run.get_output()
+        # As we have generated models, we need to register them for the future use.
+        description = "Backtest model example"
+        tags = {"last_training_date": last_training_date, "experiment": experiment.name}
+        if model_uid:
+            tags["model_uid"] = model_uid
+        automl_run.register_model(
+            model_name=best_run.properties["model_name"],
+            description=description,
+            tags=tags,
+        )
+        print(f"The model {best_run.properties['model_name']} was registered.")
+
+    _, x_pred = fitted_model.forecast(X_test)
+    x_pred.reset_index(inplace=True, drop=False)
+    columns = [automl_settings[constants.TimeSeries.TIME_COLUMN_NAME]]
+    if automl_settings.get(constants.TimeSeries.GRAIN_COLUMN_NAMES):
+        # We know that fitted_model.grain_column_names is a list.
+        columns.extend(fitted_model.grain_column_names)
+    columns.append(constants.TimeSeriesInternal.DUMMY_TARGET_COLUMN)
+    # Remove featurized columns.
+    x_pred = x_pred[columns]
+    x_pred.rename(
+        {constants.TimeSeriesInternal.DUMMY_TARGET_COLUMN: "predicted_level"},
+        axis=1,
+        inplace=True,
+    )
+    x_pred["actual_level"] = y_test
+    x_pred["backtest_iteration"] = f"iteration_{last_training_date}"
+    date_safe = RE_INVALID_SYMBOLS.sub("_", last_training_date)
+    x_pred.to_csv(os.path.join(output_dir, f"iteration_{date_safe}.csv"), index=False)
+    return x_pred
+
+
+def run(input_files):
+    """Run the script"""
+    logger.info("Running mini batch.")
+    ws = get_run().experiment.workspace
+    file_name = None
+    if model_name:
+        models = Model.list(ws, name=model_name)
+        cloud_model = None
+        if models:
+            for one_mod in models:
+                if cloud_model is None or one_mod.version > cloud_model.version:
+                    logger.info(
+                        "Using existing model from the workspace. Model version: {}".format(
+                            one_mod.version
+                        )
+                    )
+                    cloud_model = one_mod
+        file_name = cloud_model.download(exist_ok=True)
+
+    forecasts = []
+    logger.info("Running backtest.")
+    for input_file in input_files:
+        forecasts.append(run_backtest(input_file, file_name, get_run().experiment))
+    return pd.concat(forecasts)

how-to-use-azureml/automated-machine-learning/forecasting-backtest-single-model/assets/score.py

@@ -0,0 +1,167 @@
+from typing import Any, Dict, Optional, List
+
+import argparse
+import json
+import os
+import re
+
+import pandas as pd
+
+from matplotlib import pyplot as plt
+from matplotlib.backends.backend_pdf import PdfPages
+
+from azureml.automl.core.shared import constants
+from azureml.automl.core.shared.types import GrainType
+from azureml.automl.runtime.shared.score import scoring
+
+GRAIN = "time_series_id"
+BACKTEST_ITER = "backtest_iteration"
+ACTUALS = "actual_level"
+PREDICTIONS = "predicted_level"
+ALL_GRAINS = "all_sets"
+
+FORECASTS_FILE = "forecast.csv"
+SCORES_FILE = "scores.csv"
+PLOTS_FILE = "plots_fcst_vs_actual.pdf"
+RE_INVALID_SYMBOLS = re.compile("[: ]")
+
+
+def _compute_metrics(df: pd.DataFrame, metrics: List[str]):
+    """
+    Compute metrics for one data frame.
+
+    :param df: The data frame which contains actual_level and predicted_level columns.
+    :return: The data frame with two columns - metric_name and metric.
+    """
+    scores = scoring.score_regression(
+        y_test=df[ACTUALS], y_pred=df[PREDICTIONS], metrics=metrics
+    )
+    metrics_df = pd.DataFrame(list(scores.items()), columns=["metric_name", "metric"])
+    metrics_df.sort_values(["metric_name"], inplace=True)
+    metrics_df.reset_index(drop=True, inplace=True)
+    return metrics_df
+
+
+def _format_grain_name(grain: GrainType) -> str:
+    """
+    Convert grain name to string.
+
+    :param grain: the grain name.
+    :return: the string representation of the given grain.
+    """
+    if not isinstance(grain, tuple) and not isinstance(grain, list):
+        return str(grain)
+    grain = list(map(str, grain))
+    return "|".join(grain)
+
+
+def compute_all_metrics(
+    fcst_df: pd.DataFrame,
+    ts_id_colnames: List[str],
+    metric_names: Optional[List[set]] = None,
+):
+    """
+    Calculate metrics per grain.
+
+    :param fcst_df: forecast data frame. Must contain 2 columns: 'actual_level' and 'predicted_level'
+    :param metric_names: (optional) the list of metric names to return
+    :param ts_id_colnames: (optional) list of grain column names
+    :return: dictionary of summary table for all tests and final decision on stationary vs nonstaionary
+    """
+    if not metric_names:
+        metric_names = list(constants.Metric.SCALAR_REGRESSION_SET)
+
+    if ts_id_colnames is None:
+        ts_id_colnames = []
+
+    metrics_list = []
+    if ts_id_colnames:
+        for grain, df in fcst_df.groupby(ts_id_colnames):
+            one_grain_metrics_df = _compute_metrics(df, metric_names)
+            one_grain_metrics_df[GRAIN] = _format_grain_name(grain)
+            metrics_list.append(one_grain_metrics_df)
+
+    # overall metrics
+    one_grain_metrics_df = _compute_metrics(fcst_df, metric_names)
+    one_grain_metrics_df[GRAIN] = ALL_GRAINS
+    metrics_list.append(one_grain_metrics_df)
+
+    # collect into a data frame
+    return pd.concat(metrics_list)
+
+
+def _draw_one_plot(
+    df: pd.DataFrame,
+    time_column_name: str,
+    grain_column_names: List[str],
+    pdf: PdfPages,
+) -> None:
+    """
+    Draw the single plot.
+
+    :param df: The data frame with the data to build plot.
+    :param time_column_name: The name of a time column.
+    :param grain_column_names: The name of grain columns.
+    :param pdf: The pdf backend used to render the plot.
+    """
+    fig, _ = plt.subplots(figsize=(20, 10))
+    df = df.set_index(time_column_name)
+    plt.plot(df[[ACTUALS, PREDICTIONS]])
+    plt.xticks(rotation=45)
+    iteration = df[BACKTEST_ITER].iloc[0]
+    if grain_column_names:
+        grain_name = [df[grain].iloc[0] for grain in grain_column_names]
+        plt.title(f"Time series ID: {_format_grain_name(grain_name)} {iteration}")
+    plt.legend(["actual", "forecast"])
+    plt.close(fig)
+    pdf.savefig(fig)
+
+
+def calculate_scores_and_build_plots(
+    input_dir: str, output_dir: str, automl_settings: Dict[str, Any]
+):
+    os.makedirs(output_dir, exist_ok=True)
+    grains = automl_settings.get(constants.TimeSeries.GRAIN_COLUMN_NAMES)
+    time_column_name = automl_settings.get(constants.TimeSeries.TIME_COLUMN_NAME)
+    if grains is None:
+        grains = []
+    if isinstance(grains, str):
+        grains = [grains]
+    while BACKTEST_ITER in grains:
+        grains.remove(BACKTEST_ITER)
+
+    dfs = []
+    for fle in os.listdir(input_dir):
+        file_path = os.path.join(input_dir, fle)
+        if os.path.isfile(file_path) and file_path.endswith(".csv"):
+            df_iter = pd.read_csv(file_path, parse_dates=[time_column_name])
+            for _, iteration in df_iter.groupby(BACKTEST_ITER):
+                dfs.append(iteration)
+    forecast_df = pd.concat(dfs, sort=False, ignore_index=True)
+    # To make sure plots are in order, sort the predictions by grain and iteration.
+    ts_index = grains + [BACKTEST_ITER]
+    forecast_df.sort_values(by=ts_index, inplace=True)
+    pdf = PdfPages(os.path.join(output_dir, PLOTS_FILE))
+    for _, one_forecast in forecast_df.groupby(ts_index):
+        _draw_one_plot(one_forecast, time_column_name, grains, pdf)
+    pdf.close()
+    forecast_df.to_csv(os.path.join(output_dir, FORECASTS_FILE), index=False)
+    metrics = compute_all_metrics(forecast_df, grains + [BACKTEST_ITER])
+    metrics.to_csv(os.path.join(output_dir, SCORES_FILE), index=False)
+
+
+if __name__ == "__main__":
+    args = {"forecasts": "--forecasts", "scores_out": "--output-dir"}
+    parser = argparse.ArgumentParser("Parsing input arguments.")
+    for argname, arg in args.items():
+        parser.add_argument(arg, dest=argname, required=True)
+    parsed_args, _ = parser.parse_known_args()
+    input_dir = parsed_args.forecasts
+    output_dir = parsed_args.scores_out
+    with open(
+        os.path.join(
+            os.path.dirname(os.path.realpath(__file__)), "automl_settings.json"
+        )
+    ) as json_file:
+        automl_settings = json.load(json_file)
+    calculate_scores_and_build_plots(input_dir, output_dir, automl_settings)

how-to-use-azureml/automated-machine-learning/forecasting-backtest-single-model/auto-ml-forecasting-backtest-single-model.ipynb

@@ -0,0 +1,719 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Copyright (c) Microsoft Corporation. All rights reserved.\n",
+    "\n",
+    "Licensed under the MIT License.\n",
+    "![Impressions](https://PixelServer20190423114238.azurewebsites.net/api/impressions/MachineLearningNotebooks/how-to-use-azureml/automated-machine-learning/automl-forecasting-function.png)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Automated MachineLearning\n",
+    "_**The model backtesting**_\n",
+    "\n",
+    "## Contents\n",
+    "1. [Introduction](#Introduction)\n",
+    "2. [Setup](#Setup)\n",
+    "3. [Data](#Data)\n",
+    "4. [Prepare remote compute and data.](#prepare_remote)\n",
+    "5. [Create the configuration for AutoML backtesting](#train)\n",
+    "6. [Backtest AutoML](#backtest_automl)\n",
+    "7. [View metrics](#Metrics)\n",
+    "8. [Backtest the best model](#backtest_model)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Introduction\n",
+    "Model backtesting is used to evaluate its performance on historical data. To do that we step back on the backtesting period by the data set several times and split the data to train and test sets. Then these data sets are used for training and evaluation of model.<br>\n",
+    "This notebook is intended to demonstrate backtesting on a single model, this is the best solution for small data sets with a few or one time series in it. For scenarios where we would like to choose the best AutoML model for every backtest iteration, please see [AutoML Forecasting Backtest Many Models Example](../forecasting-backtest-many-models/auto-ml-forecasting-backtest-many-models.ipynb) notebook.\n",
+    "![Backtesting](Backtesting.png)\n",
+    "This notebook demonstrates two ways of backtesting:\n",
+    "- AutoML backtesting: we will train separate AutoML models for historical data\n",
+    "- Model backtesting: from the first run we will select the best model trained on the most recent data, retrain it on the past data and evaluate."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Setup"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "import numpy as np\n",
+    "import pandas as pd\n",
+    "import shutil\n",
+    "\n",
+    "import azureml.core\n",
+    "from azureml.core import Experiment, Model, Workspace"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "This notebook is compatible with Azure ML SDK version 1.35.1 or later."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "print(\"You are currently using version\", azureml.core.VERSION, \"of the Azure ML SDK\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "As part of the setup you have already created a <b>Workspace</b>."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "ws = Workspace.from_config()\n",
+    "\n",
+    "output = {}\n",
+    "output[\"Subscription ID\"] = ws.subscription_id\n",
+    "output[\"Workspace\"] = ws.name\n",
+    "output[\"SKU\"] = ws.sku\n",
+    "output[\"Resource Group\"] = ws.resource_group\n",
+    "output[\"Location\"] = ws.location\n",
+    "pd.set_option(\"display.max_colwidth\", -1)\n",
+    "outputDf = pd.DataFrame(data=output, index=[\"\"])\n",
+    "outputDf.T"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Data\n",
+    "For the demonstration purposes we will simulate one year of daily data. To do this we need to specify the following parameters: time column name, time series ID column names and label column name. Our intention is to forecast for two weeks ahead."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "TIME_COLUMN_NAME = \"date\"\n",
+    "TIME_SERIES_ID_COLUMN_NAMES = \"time_series_id\"\n",
+    "LABEL_COLUMN_NAME = \"y\"\n",
+    "FORECAST_HORIZON = 14\n",
+    "FREQUENCY = \"D\"\n",
+    "\n",
+    "\n",
+    "def simulate_timeseries_data(\n",
+    "    train_len: int,\n",
+    "    test_len: int,\n",
+    "    time_column_name: str,\n",
+    "    target_column_name: str,\n",
+    "    time_series_id_column_name: str,\n",
+    "    time_series_number: int = 1,\n",
+    "    freq: str = \"H\",\n",
+    "):\n",
+    "    \"\"\"\n",
+    "    Return the time series of designed length.\n",
+    "\n",
+    "    :param train_len: The length of training data (one series).\n",
+    "    :type train_len: int\n",
+    "    :param test_len: The length of testing data (one series).\n",
+    "    :type test_len: int\n",
+    "    :param time_column_name: The desired name of a time column.\n",
+    "    :type time_column_name: str\n",
+    "    :param time_series_number: The number of time series in the data set.\n",
+    "    :type time_series_number: int\n",
+    "    :param freq: The frequency string representing pandas offset.\n",
+    "                 see https://pandas.pydata.org/pandas-docs/stable/user_guide/timeseries.html\n",
+    "    :type freq: str\n",
+    "    :returns: the tuple of train and test data sets.\n",
+    "    :rtype: tuple\n",
+    "\n",
+    "    \"\"\"\n",
+    "    data_train = []  # type: List[pd.DataFrame]\n",
+    "    data_test = []  # type: List[pd.DataFrame]\n",
+    "    data_length = train_len + test_len\n",
+    "    for i in range(time_series_number):\n",
+    "        X = pd.DataFrame(\n",
+    "            {\n",
+    "                time_column_name: pd.date_range(\n",
+    "                    start=\"2000-01-01\", periods=data_length, freq=freq\n",
+    "                ),\n",
+    "                target_column_name: np.arange(data_length).astype(float)\n",
+    "                + np.random.rand(data_length)\n",
+    "                + i * 5,\n",
+    "                \"ext_predictor\": np.asarray(range(42, 42 + data_length)),\n",
+    "                time_series_id_column_name: np.repeat(\"ts{}\".format(i), data_length),\n",
+    "            }\n",
+    "        )\n",
+    "        data_train.append(X[:train_len])\n",
+    "        data_test.append(X[train_len:])\n",
+    "    train = pd.concat(data_train)\n",
+    "    label_train = train.pop(target_column_name).values\n",
+    "    test = pd.concat(data_test)\n",
+    "    label_test = test.pop(target_column_name).values\n",
+    "    return train, label_train, test, label_test\n",
+    "\n",
+    "\n",
+    "n_test_periods = FORECAST_HORIZON\n",
+    "n_train_periods = 365\n",
+    "X_train, y_train, X_test, y_test = simulate_timeseries_data(\n",
+    "    train_len=n_train_periods,\n",
+    "    test_len=n_test_periods,\n",
+    "    time_column_name=TIME_COLUMN_NAME,\n",
+    "    target_column_name=LABEL_COLUMN_NAME,\n",
+    "    time_series_id_column_name=TIME_SERIES_ID_COLUMN_NAMES,\n",
+    "    time_series_number=2,\n",
+    "    freq=FREQUENCY,\n",
+    ")\n",
+    "X_train[LABEL_COLUMN_NAME] = y_train"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Let's see what the training data looks like."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "X_train.tail()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Prepare remote compute and data. <a id=\"prepare_remote\"></a>\n",
+    "The [Machine Learning service workspace](https://docs.microsoft.com/en-us/azure/machine-learning/service/concept-workspace), is paired with the storage account, which contains the default data store. We will use it to upload the artificial data and create [tabular dataset](https://docs.microsoft.com/en-us/python/api/azureml-core/azureml.data.tabulardataset?view=azure-ml-py) for training. A tabular dataset defines a series of lazily-evaluated, immutable operations to load data from the data source into tabular representation."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.data.dataset_factory import TabularDatasetFactory\n",
+    "\n",
+    "ds = ws.get_default_datastore()\n",
+    "# Upload saved data to the default data store.\n",
+    "train_data = TabularDatasetFactory.register_pandas_dataframe(\n",
+    "    X_train, target=(ds, \"data\"), name=\"data_backtest\"\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "You will need to create a compute target for backtesting. In this [tutorial](https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-set-up-training-targets#amlcompute), you create AmlCompute as your training compute resource.\n",
+    "\n",
+    "> Note that if you have an AzureML Data Scientist role, you will not have permission to create compute resources. Talk to your workspace or IT admin to create the compute targets described in this section, if they do not already exist."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.core.compute import ComputeTarget, AmlCompute\n",
+    "from azureml.core.compute_target import ComputeTargetException\n",
+    "\n",
+    "# Choose a name for your CPU cluster\n",
+    "amlcompute_cluster_name = \"backtest-cluster\"\n",
+    "\n",
+    "# Verify that cluster does not exist already\n",
+    "try:\n",
+    "    compute_target = ComputeTarget(workspace=ws, name=amlcompute_cluster_name)\n",
+    "    print(\"Found existing cluster, use it.\")\n",
+    "except ComputeTargetException:\n",
+    "    compute_config = AmlCompute.provisioning_configuration(\n",
+    "        vm_size=\"STANDARD_DS12_V2\", max_nodes=6\n",
+    "    )\n",
+    "    compute_target = ComputeTarget.create(ws, amlcompute_cluster_name, compute_config)\n",
+    "\n",
+    "compute_target.wait_for_completion(show_output=True)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Create the configuration for AutoML backtesting <a id=\"train\"></a>\n",
+    "\n",
+    "This dictionary defines the AutoML and many models settings. For this forecasting task we need to define several settings including the name of the time column, the maximum forecast horizon, and the partition column name definition.\n",
+    "\n",
+    "| Property                           | Description|\n",
+    "| :---------------                   | :------------------- |\n",
+    "| **task**                           | forecasting |\n",
+    "| **primary_metric**                 | This is the metric that you want to optimize.<br> Forecasting supports the following primary metrics <br><i>normalized_root_mean_squared_error</i><br><i>normalized_mean_absolute_error</i> |\n",
+    "| **iteration_timeout_minutes**      | Maximum amount of time in minutes that the model can train. This is optional but provides customers with greater control on exit criteria. |\n",
+    "| **iterations**                     | Number of models to train. This is optional but provides customers with greater control on exit criteria. |\n",
+    "| **experiment_timeout_hours**       | Maximum amount of time in hours that the experiment can take before it terminates. This is optional but provides customers with greater control on exit criteria. |\n",
+    "| **label_column_name**              | The name of the label column. |\n",
+    "| **max_horizon**               | The forecast horizon is how many periods forward you would like to forecast. This integer horizon is in units of the timeseries frequency (e.g. daily, weekly). Periods are inferred from your data. |\n",
+    "| **n_cross_validations**            | Number of cross validation splits. Rolling Origin Validation is used to split time-series in a temporally consistent way. |\n",
+    "| **time_column_name**               | The name of your time column. |\n",
+    "| **grain_column_names**     | The column names used to uniquely identify timeseries in data that has multiple rows with the same timestamp. |"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "automl_settings = {\n",
+    "    \"task\": \"forecasting\",\n",
+    "    \"primary_metric\": \"normalized_root_mean_squared_error\",\n",
+    "    \"iteration_timeout_minutes\": 10,  # This needs to be changed based on the dataset. We ask customer to explore how long training is taking before settings this value\n",
+    "    \"iterations\": 15,\n",
+    "    \"experiment_timeout_hours\": 1,  # This also needs to be changed based on the dataset. For larger data set this number needs to be bigger.\n",
+    "    \"label_column_name\": LABEL_COLUMN_NAME,\n",
+    "    \"n_cross_validations\": 3,\n",
+    "    \"time_column_name\": TIME_COLUMN_NAME,\n",
+    "    \"max_horizon\": FORECAST_HORIZON,\n",
+    "    \"track_child_runs\": False,\n",
+    "    \"grain_column_names\": TIME_SERIES_ID_COLUMN_NAMES,\n",
+    "}"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Backtest AutoML <a id=\"backtest_automl\"></a>\n",
+    "First we set backtesting parameters: we will step back by 30 days and will make 5 such steps; for each step we will forecast for next two weeks."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# The number of periods to step back on each backtest iteration.\n",
+    "BACKTESTING_PERIOD = 30\n",
+    "# The number of times we will back test the model.\n",
+    "NUMBER_OF_BACKTESTS = 5"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "To train AutoML on backtesting folds we will use the [Azure Machine Learning pipeline](https://docs.microsoft.com/en-us/azure/machine-learning/concept-ml-pipelines). It will generate backtest folds, then train model for each of them and calculate the accuracy metrics. To run pipeline, you also need to create an <b>Experiment</b>. An Experiment corresponds to a prediction problem you are trying to solve (here, it is a forecasting), while a Run corresponds to a specific approach to the problem."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from uuid import uuid1\n",
+    "\n",
+    "from pipeline_helper import get_backtest_pipeline\n",
+    "\n",
+    "pipeline_exp = Experiment(ws, \"automl-backtesting\")\n",
+    "\n",
+    "# We will create the unique identifier to mark our models.\n",
+    "model_uid = str(uuid1())\n",
+    "\n",
+    "pipeline = get_backtest_pipeline(\n",
+    "    experiment=pipeline_exp,\n",
+    "    dataset=train_data,\n",
+    "    # The STANDARD_DS12_V2 has 4 vCPU per node, we will set 2 process per node to be safe.\n",
+    "    process_per_node=2,\n",
+    "    # The maximum number of nodes for our compute is 6.\n",
+    "    node_count=6,\n",
+    "    compute_target=compute_target,\n",
+    "    automl_settings=automl_settings,\n",
+    "    step_size=BACKTESTING_PERIOD,\n",
+    "    step_number=NUMBER_OF_BACKTESTS,\n",
+    "    model_uid=model_uid,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Run the pipeline and wait for results."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "pipeline_run = pipeline_exp.submit(pipeline)\n",
+    "pipeline_run.wait_for_completion(show_output=False)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "After the run is complete, we can download the results. "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "metrics_output = pipeline_run.get_pipeline_output(\"results\")\n",
+    "metrics_output.download(\"backtest_metrics\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## View metrics<a id=\"Metrics\"></a>\n",
+    "To distinguish these metrics from the model backtest, which we will obtain in the next section, we will move the directory with metrics out of the backtest_metrics and will remove the parent folder. We will create the utility function for that."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def copy_scoring_directory(new_name):\n",
+    "    scores_path = os.path.join(\"backtest_metrics\", \"azureml\")\n",
+    "    directory_list = [os.path.join(scores_path, d) for d in os.listdir(scores_path)]\n",
+    "    latest_file = max(directory_list, key=os.path.getctime)\n",
+    "    print(\n",
+    "        f\"The output directory {latest_file} was created on {pd.Timestamp(os.path.getctime(latest_file), unit='s')} GMT.\"\n",
+    "    )\n",
+    "    shutil.move(os.path.join(latest_file, \"results\"), new_name)\n",
+    "    shutil.rmtree(\"backtest_metrics\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Move the directory and list its contents."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "copy_scoring_directory(\"automl_backtest\")\n",
+    "pd.DataFrame({\"File\": os.listdir(\"automl_backtest\")})"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "The directory contains a set of files with results:\n",
+    "- forecast.csv contains forecasts for all backtest iterations. The backtest_iteration column contains iteration identifier with the last training date as a suffix\n",
+    "- scores.csv contains all metrics. If data set contains several time series, the metrics are given for all combinations of time series id and iterations, as well as scores for all iterations and time series id are marked as \"all_sets\"\n",
+    "- plots_fcst_vs_actual.pdf contains the predictions vs forecast plots for each iteration and time series.\n",
+    "\n",
+    "For demonstration purposes we will display the table of metrics for one of the time series with ID \"ts0\". Again, we will create the utility function, which will be re used in model backtesting."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def get_metrics_for_ts(all_metrics, ts):\n",
+    "    \"\"\"\n",
+    "    Get the metrics for the time series with ID ts and return it as pandas data frame.\n",
+    "\n",
+    "    :param all_metrics: The table with all the metrics.\n",
+    "    :param ts: The ID of a time series of interest.\n",
+    "    :return: The pandas DataFrame with metrics for one time series.\n",
+    "    \"\"\"\n",
+    "    results_df = None\n",
+    "    for ts_id, one_series in all_metrics.groupby(\"time_series_id\"):\n",
+    "        if not ts_id.startswith(ts):\n",
+    "            continue\n",
+    "        iteration = ts_id.split(\"|\")[-1]\n",
+    "        df = one_series[[\"metric_name\", \"metric\"]]\n",
+    "        df.rename({\"metric\": iteration}, axis=1, inplace=True)\n",
+    "        df.set_index(\"metric_name\", inplace=True)\n",
+    "        if results_df is None:\n",
+    "            results_df = df\n",
+    "        else:\n",
+    "            results_df = results_df.merge(\n",
+    "                df, how=\"inner\", left_index=True, right_index=True\n",
+    "            )\n",
+    "    results_df.sort_index(axis=1, inplace=True)\n",
+    "    return results_df\n",
+    "\n",
+    "\n",
+    "metrics_df = pd.read_csv(os.path.join(\"automl_backtest\", \"scores.csv\"))\n",
+    "ts_id = \"ts0\"\n",
+    "get_metrics_for_ts(metrics_df, ts_id)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Forecast vs actuals plots."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from IPython.display import IFrame\n",
+    "\n",
+    "IFrame(\"./automl_backtest/plots_fcst_vs_actual.pdf\", width=800, height=300)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# <font color='blue'>Backtest the best model</font> <a id=\"backtest_model\"></a>\n",
+    "\n",
+    "For model backtesting we will use the same parameters we used to backtest AutoML. All the models, we have obtained in the previous run were registered in our workspace. To identify the model, each was assigned a tag with the last trainig date."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model_list = Model.list(ws, tags={\"experiment\": \"automl-backtesting\"})\n",
+    "model_data = {\"name\": [], \"last_training_date\": []}\n",
+    "for model in model_list:\n",
+    "    if (\n",
+    "        \"last_training_date\" not in model.tags\n",
+    "        or \"model_uid\" not in model.tags\n",
+    "        or model.tags[\"model_uid\"] != model_uid\n",
+    "    ):\n",
+    "        continue\n",
+    "    model_data[\"name\"].append(model.name)\n",
+    "    model_data[\"last_training_date\"].append(\n",
+    "        pd.Timestamp(model.tags[\"last_training_date\"])\n",
+    "    )\n",
+    "df_models = pd.DataFrame(model_data)\n",
+    "df_models.sort_values([\"last_training_date\"], inplace=True)\n",
+    "df_models.reset_index(inplace=True, drop=True)\n",
+    "df_models"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "We will backtest the model trained on the most recet data."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model_name = df_models[\"name\"].iloc[-1]\n",
+    "model_name"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Retrain the models.\n",
+    "Assemble the pipeline, which will retrain the best model from AutoML run on historical data."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "pipeline_exp = Experiment(ws, \"model-backtesting\")\n",
+    "\n",
+    "pipeline = get_backtest_pipeline(\n",
+    "    experiment=pipeline_exp,\n",
+    "    dataset=train_data,\n",
+    "    # The STANDARD_DS12_V2 has 4 vCPU per node, we will set 2 process per node to be safe.\n",
+    "    process_per_node=2,\n",
+    "    # The maximum number of nodes for our compute is 6.\n",
+    "    node_count=6,\n",
+    "    compute_target=compute_target,\n",
+    "    automl_settings=automl_settings,\n",
+    "    step_size=BACKTESTING_PERIOD,\n",
+    "    step_number=NUMBER_OF_BACKTESTS,\n",
+    "    model_name=model_name,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Launch the backtesting pipeline."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "pipeline_run = pipeline_exp.submit(pipeline)\n",
+    "pipeline_run.wait_for_completion(show_output=False)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "The metrics are stored in the pipeline output named \"score\". The next code will download the table with metrics."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "metrics_output = pipeline_run.get_pipeline_output(\"results\")\n",
+    "metrics_output.download(\"backtest_metrics\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Again, we will copy the data files from the downloaded directory, but in this case we will call the folder \"model_backtest\"; it will contain the same files as the one for AutoML backtesting."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "copy_scoring_directory(\"model_backtest\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Finally, we will display the metrics."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model_metrics_df = pd.read_csv(os.path.join(\"model_backtest\", \"scores.csv\"))\n",
+    "get_metrics_for_ts(model_metrics_df, ts_id)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Forecast vs actuals plots."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from IPython.display import IFrame\n",
+    "\n",
+    "IFrame(\"./model_backtest/plots_fcst_vs_actual.pdf\", width=800, height=300)"
+   ]
+  }
+ ],
+ "metadata": {
+  "authors": [
+   {
+    "name": "jialiu"
+   }
+  ],
+  "category": "tutorial",
+  "compute": [
+   "Remote"
+  ],
+  "datasets": [
+   "None"
+  ],
+  "deployment": [
+   "None"
+  ],
+  "exclude_from_index": false,
+  "framework": [
+   "Azure ML AutoML"
+  ],
+  "kernelspec": {
+   "display_name": "Python 3.6 - AzureML",
+   "language": "python",
+   "name": "python3-azureml"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.6.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}

how-to-use-azureml/automated-machine-learning/forecasting-backtest-single-model/auto-ml-forecasting-backtest-single-model.yml

@@ -0,0 +1,4 @@
+name: auto-ml-forecasting-backtest-single-model
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/forecasting-backtest-single-model/pipeline_helper.py

@@ -0,0 +1,166 @@
+from typing import Any, Dict, Optional
+
+import os
+
+import azureml.train.automl.runtime._hts.hts_runtime_utilities as hru
+
+from azureml._restclient.jasmine_client import JasmineClient
+from azureml.contrib.automl.pipeline.steps import utilities
+from azureml.core import RunConfiguration
+from azureml.core.compute import ComputeTarget
+from azureml.core.experiment import Experiment
+from azureml.data import LinkTabularOutputDatasetConfig, TabularDataset
+from azureml.pipeline.core import Pipeline, PipelineData, PipelineParameter
+from azureml.pipeline.steps import ParallelRunConfig, ParallelRunStep, PythonScriptStep
+from azureml.train.automl.constants import Scenarios
+from azureml.data.dataset_consumption_config import DatasetConsumptionConfig
+
+
+PROJECT_FOLDER = "assets"
+SETTINGS_FILE = "automl_settings.json"
+
+
+def get_backtest_pipeline(
+    experiment: Experiment,
+    dataset: TabularDataset,
+    process_per_node: int,
+    node_count: int,
+    compute_target: ComputeTarget,
+    automl_settings: Dict[str, Any],
+    step_size: int,
+    step_number: int,
+    model_name: Optional[str] = None,
+    model_uid: Optional[str] = None,
+) -> Pipeline:
+    """
+    :param experiment: The experiment used to run the pipeline.
+    :param dataset: Tabular data set to be used for model training.
+    :param process_per_node: The number of processes per node. Generally it should be the number of cores
+                             on the node divided by two.
+    :param node_count: The number of nodes to be used.
+    :param compute_target: The compute target to be used to run the pipeline.
+    :param model_name: The name of a model to be back tested.
+    :param automl_settings: The dictionary with automl settings.
+    :param step_size: The number of periods to step back in backtesting.
+    :param step_number: The number of backtesting iterations.
+    :param model_uid: The uid to mark models from this run of the experiment.
+    :return: The pipeline to be used for model retraining.
+             **Note:** The output will be uploaded in the pipeline output
+             called 'score'.
+    """
+    jasmine_client = JasmineClient(
+        service_context=experiment.workspace.service_context,
+        experiment_name=experiment.name,
+        experiment_id=experiment.id,
+    )
+    env = jasmine_client.get_curated_environment(
+        scenario=Scenarios.AUTOML,
+        enable_dnn=False,
+        enable_gpu=False,
+        compute=compute_target,
+        compute_sku=experiment.workspace.compute_targets.get(
+            compute_target.name
+        ).vm_size,
+    )
+    data_results = PipelineData(
+        name="results", datastore=None, pipeline_output_name="results"
+    )
+    ############################################################
+    # Split the data set using python script.
+    ############################################################
+    run_config = RunConfiguration()
+    run_config.docker.use_docker = True
+    run_config.environment = env
+
+    split_data = PipelineData(name="split_data_output", datastore=None).as_dataset()
+    split_step = PythonScriptStep(
+        name="split_data_for_backtest",
+        script_name="data_split.py",
+        inputs=[dataset.as_named_input("training_data")],
+        outputs=[split_data],
+        source_directory=PROJECT_FOLDER,
+        arguments=[
+            "--step-size",
+            step_size,
+            "--step-number",
+            step_number,
+            "--time-column-name",
+            automl_settings.get("time_column_name"),
+            "--time-series-id-column-names",
+            automl_settings.get("grain_column_names"),
+            "--output-dir",
+            split_data,
+        ],
+        runconfig=run_config,
+        compute_target=compute_target,
+        allow_reuse=False,
+    )
+    ############################################################
+    # We will do the backtest the parallel run step.
+    ############################################################
+    settings_path = os.path.join(PROJECT_FOLDER, SETTINGS_FILE)
+    hru.dump_object_to_json(automl_settings, settings_path)
+    mini_batch_size = PipelineParameter(name="batch_size_param", default_value=str(1))
+    back_test_config = ParallelRunConfig(
+        source_directory=PROJECT_FOLDER,
+        entry_script="retrain_models.py",
+        mini_batch_size=mini_batch_size,
+        error_threshold=-1,
+        output_action="append_row",
+        append_row_file_name="outputs.txt",
+        compute_target=compute_target,
+        environment=env,
+        process_count_per_node=process_per_node,
+        run_invocation_timeout=3600,
+        node_count=node_count,
+    )
+    forecasts = PipelineData(name="forecasts", datastore=None)
+    if model_name:
+        parallel_step_name = "{}-backtest".format(model_name.replace("_", "-"))
+    else:
+        parallel_step_name = "AutoML-backtest"
+
+    prs_args = [
+        "--target_column_name",
+        automl_settings.get("label_column_name"),
+        "--output-dir",
+        forecasts,
+    ]
+    if model_name is not None:
+        prs_args.append("--model-name")
+        prs_args.append(model_name)
+    if model_uid is not None:
+        prs_args.append("--model-uid")
+        prs_args.append(model_uid)
+    backtest_prs = ParallelRunStep(
+        name=parallel_step_name,
+        parallel_run_config=back_test_config,
+        arguments=prs_args,
+        inputs=[split_data],
+        output=forecasts,
+        allow_reuse=False,
+    )
+    ############################################################
+    # Then we collect the output and return it as scores output.
+    ############################################################
+    collection_step = PythonScriptStep(
+        name="score",
+        script_name="score.py",
+        inputs=[forecasts.as_mount()],
+        outputs=[data_results],
+        source_directory=PROJECT_FOLDER,
+        arguments=[
+            "--forecasts",
+            forecasts,
+            "--output-dir",
+            data_results,
+        ],
+        runconfig=run_config,
+        compute_target=compute_target,
+        allow_reuse=False,
+    )
+    # Build and return the pipeline.
+    return Pipeline(
+        workspace=experiment.workspace,
+        steps=[split_step, backtest_prs, collection_step],
+    )

how-to-use-azureml/automated-machine-learning/forecasting-beer-remote/auto-ml-forecasting-beer-remote.ipynb

@@ -113,7 +113,7 @@
       "metadata": {},
       "outputs": [],
       "source": [
-        "print(\"This notebook was created using version 1.23.0 of the Azure ML SDK\")\n",
+        "print(\"This notebook was created using version 1.38.0 of the Azure ML SDK\")\n",
         "print(\"You are currently using version\", azureml.core.VERSION, \"of the Azure ML SDK\")"
       ]
     },
@@ -139,18 +139,18 @@
         "ws = Workspace.from_config()\n",
         "\n",
         "# choose a name for the run history container in the workspace\n",
-        "experiment_name = 'beer-remote-cpu'\n",
+        "experiment_name = \"beer-remote-cpu\"\n",
         "\n",
         "experiment = Experiment(ws, experiment_name)\n",
         "\n",
         "output = {}\n",
-        "output['Subscription ID'] = ws.subscription_id\n",
-        "output['Workspace'] = ws.name\n",
-        "output['Resource Group'] = ws.resource_group\n",
-        "output['Location'] = ws.location\n",
-        "output['Run History Name'] = experiment_name\n",
-        "pd.set_option('display.max_colwidth', -1)\n",
-        "outputDf = pd.DataFrame(data = output, index = [''])\n",
+        "output[\"Subscription ID\"] = ws.subscription_id\n",
+        "output[\"Workspace\"] = ws.name\n",
+        "output[\"Resource Group\"] = ws.resource_group\n",
+        "output[\"Location\"] = ws.location\n",
+        "output[\"Run History Name\"] = experiment_name\n",
+        "pd.set_option(\"display.max_colwidth\", -1)\n",
+        "outputDf = pd.DataFrame(data=output, index=[\"\"])\n",
         "outputDf.T"
       ]
     },
@@ -162,7 +162,9 @@
       },
       "source": [
         "### Using AmlCompute\n",
-        "You will need to create a [compute target](https://docs.microsoft.com/azure/machine-learning/service/concept-azure-machine-learning-architecture#compute-target) for your AutoML run. In this tutorial, you use `AmlCompute` as your training compute resource."
+        "You will need to create a [compute target](https://docs.microsoft.com/azure/machine-learning/service/concept-azure-machine-learning-architecture#compute-target) for your AutoML run. In this tutorial, you use `AmlCompute` as your training compute resource.\n",
+        "\n",
+        "> Note that if you have an AzureML Data Scientist role, you will not have permission to create compute resources. Talk to your workspace or IT admin to create the compute targets described in this section, if they do not already exist."
       ]
     },
     {
@@ -183,10 +185,11 @@
         "# Verify that cluster does not exist already\n",
         "try:\n",
         "    compute_target = ComputeTarget(workspace=ws, name=cpu_cluster_name)\n",
-        "    print('Found existing cluster, use it.')\n",
+        "    print(\"Found existing cluster, use it.\")\n",
         "except ComputeTargetException:\n",
-        "    compute_config = AmlCompute.provisioning_configuration(vm_size='STANDARD_D2_V2',\n",
-        "                                                           max_nodes=4)\n",
+        "    compute_config = AmlCompute.provisioning_configuration(\n",
+        "        vm_size=\"STANDARD_DS12_V2\", max_nodes=4\n",
+        "    )\n",
         "    compute_target = ComputeTarget.create(ws, cpu_cluster_name, compute_config)\n",
         "\n",
         "compute_target.wait_for_completion(show_output=True)"
@@ -243,17 +246,21 @@
         "plt.tight_layout()\n",
         "\n",
         "plt.subplot(2, 1, 1)\n",
-        "plt.title('Beer Production By Year')\n",
-        "df = pd.read_csv(\"Beer_no_valid_split_train.csv\", parse_dates=True, index_col= 'DATE').drop(columns='grain')\n",
-        "test_df = pd.read_csv(\"Beer_no_valid_split_test.csv\", parse_dates=True, index_col= 'DATE').drop(columns='grain')\n",
+        "plt.title(\"Beer Production By Year\")\n",
+        "df = pd.read_csv(\n",
+        "    \"Beer_no_valid_split_train.csv\", parse_dates=True, index_col=\"DATE\"\n",
+        ").drop(columns=\"grain\")\n",
+        "test_df = pd.read_csv(\n",
+        "    \"Beer_no_valid_split_test.csv\", parse_dates=True, index_col=\"DATE\"\n",
+        ").drop(columns=\"grain\")\n",
         "plt.plot(df)\n",
         "\n",
         "plt.subplot(2, 1, 2)\n",
-        "plt.title('Beer Production By Month')\n",
+        "plt.title(\"Beer Production By Month\")\n",
         "groups = df.groupby(df.index.month)\n",
         "months = concat([DataFrame(x[1].values) for x in groups], axis=1)\n",
         "months = DataFrame(months)\n",
-        "months.columns = range(1,13)\n",
+        "months.columns = range(1, 13)\n",
         "months.boxplot()\n",
         "\n",
         "plt.show()"
@@ -268,10 +275,10 @@
       },
       "outputs": [],
       "source": [
-        "target_column_name = 'BeerProduction'\n",
-        "time_column_name = 'DATE'\n",
+        "target_column_name = \"BeerProduction\"\n",
+        "time_column_name = \"DATE\"\n",
         "time_series_id_column_names = []\n",
-        "freq = 'M' #Monthly data"
+        "freq = \"M\"  # Monthly data"
       ]
     },
     {
@@ -299,14 +306,36 @@
         "test_df.to_csv(\"test.csv\")\n",
         "\n",
         "datastore = ws.get_default_datastore()\n",
-        "datastore.upload_files(files = ['./train.csv'], target_path = 'beer-dataset/tabular/', overwrite = True,show_progress = True)\n",
-        "datastore.upload_files(files = ['./valid.csv'], target_path = 'beer-dataset/tabular/', overwrite = True,show_progress = True)\n",
-        "datastore.upload_files(files = ['./test.csv'], target_path = 'beer-dataset/tabular/', overwrite = True,show_progress = True)\n",
+        "datastore.upload_files(\n",
+        "    files=[\"./train.csv\"],\n",
+        "    target_path=\"beer-dataset/tabular/\",\n",
+        "    overwrite=True,\n",
+        "    show_progress=True,\n",
+        ")\n",
+        "datastore.upload_files(\n",
+        "    files=[\"./valid.csv\"],\n",
+        "    target_path=\"beer-dataset/tabular/\",\n",
+        "    overwrite=True,\n",
+        "    show_progress=True,\n",
+        ")\n",
+        "datastore.upload_files(\n",
+        "    files=[\"./test.csv\"],\n",
+        "    target_path=\"beer-dataset/tabular/\",\n",
+        "    overwrite=True,\n",
+        "    show_progress=True,\n",
+        ")\n",
         "\n",
         "from azureml.core import Dataset\n",
-        "train_dataset = Dataset.Tabular.from_delimited_files(path = [(datastore, 'beer-dataset/tabular/train.csv')])\n",
-        "valid_dataset = Dataset.Tabular.from_delimited_files(path = [(datastore, 'beer-dataset/tabular/valid.csv')])\n",
-        "test_dataset = Dataset.Tabular.from_delimited_files(path = [(datastore, 'beer-dataset/tabular/test.csv')])"
+        "\n",
+        "train_dataset = Dataset.Tabular.from_delimited_files(\n",
+        "    path=[(datastore, \"beer-dataset/tabular/train.csv\")]\n",
+        ")\n",
+        "valid_dataset = Dataset.Tabular.from_delimited_files(\n",
+        "    path=[(datastore, \"beer-dataset/tabular/valid.csv\")]\n",
+        ")\n",
+        "test_dataset = Dataset.Tabular.from_delimited_files(\n",
+        "    path=[(datastore, \"beer-dataset/tabular/test.csv\")]\n",
+        ")"
       ]
     },
     {
@@ -364,22 +393,29 @@
       "outputs": [],
       "source": [
         "from azureml.automl.core.forecasting_parameters import ForecastingParameters\n",
+        "\n",
         "forecasting_parameters = ForecastingParameters(\n",
-        "    time_column_name=time_column_name, forecast_horizon=forecast_horizon\n",
+        "    time_column_name=time_column_name,\n",
+        "    forecast_horizon=forecast_horizon,\n",
+        "    freq=\"MS\",  # Set the forecast frequency to be monthly (start of the month)\n",
         ")\n",
         "\n",
-        "automl_config = AutoMLConfig(task='forecasting',                             \n",
-        "                             primary_metric='normalized_root_mean_squared_error',\n",
-        "                             experiment_timeout_hours = 1,\n",
-        "                             training_data=train_dataset,\n",
-        "                             label_column_name=target_column_name,\n",
-        "                             validation_data=valid_dataset, \n",
-        "                             verbosity=logging.INFO,\n",
-        "                             compute_target=compute_target,\n",
-        "                             max_concurrent_iterations=4,\n",
-        "                             max_cores_per_iteration=-1,\n",
-        "                             enable_dnn=True,\n",
-        "                             forecasting_parameters=forecasting_parameters)"
+        "# We will disable the enable_early_stopping flag to ensure the DNN model is recommended for demonstration purpose.\n",
+        "automl_config = AutoMLConfig(\n",
+        "    task=\"forecasting\",\n",
+        "    primary_metric=\"normalized_root_mean_squared_error\",\n",
+        "    experiment_timeout_hours=1,\n",
+        "    training_data=train_dataset,\n",
+        "    label_column_name=target_column_name,\n",
+        "    validation_data=valid_dataset,\n",
+        "    verbosity=logging.INFO,\n",
+        "    compute_target=compute_target,\n",
+        "    max_concurrent_iterations=4,\n",
+        "    max_cores_per_iteration=-1,\n",
+        "    enable_dnn=True,\n",
+        "    enable_early_stopping=False,\n",
+        "    forecasting_parameters=forecasting_parameters,\n",
+        ")"
       ]
     },
     {
@@ -401,8 +437,7 @@
       },
       "outputs": [],
       "source": [
-        "remote_run = experiment.submit(automl_config, show_output= False)\n",
-        "remote_run"
+        "remote_run = experiment.submit(automl_config, show_output=True)"
       ]
     },
     {
@@ -419,15 +454,6 @@
         "# remote_run = AutoMLRun(experiment = experiment, run_id = '<replace with your run id>')"
       ]
     },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "remote_run.wait_for_completion()"
-      ]
-    },
     {
       "cell_type": "markdown",
       "metadata": {
@@ -459,6 +485,7 @@
       "outputs": [],
       "source": [
         "from helper import get_result_df\n",
+        "\n",
         "summary_df = get_result_df(remote_run)\n",
         "summary_df"
       ]
@@ -474,11 +501,12 @@
       "source": [
         "from azureml.core.run import Run\n",
         "from azureml.widgets import RunDetails\n",
-        "forecast_model = 'TCNForecaster'\n",
-        "if not forecast_model in summary_df['run_id']:\n",
-        "    forecast_model = 'ForecastTCN'\n",
-        "    \n",
-        "best_dnn_run_id = summary_df['run_id'][forecast_model]\n",
+        "\n",
+        "forecast_model = \"TCNForecaster\"\n",
+        "if not forecast_model in summary_df[\"run_id\"]:\n",
+        "    forecast_model = \"ForecastTCN\"\n",
+        "\n",
+        "best_dnn_run_id = summary_df[\"run_id\"][forecast_model]\n",
         "best_dnn_run = Run(experiment, best_dnn_run_id)"
       ]
     },
@@ -492,7 +520,7 @@
       "outputs": [],
       "source": [
         "best_dnn_run.parent\n",
-        "RunDetails(best_dnn_run.parent).show() "
+        "RunDetails(best_dnn_run.parent).show()"
       ]
     },
     {
@@ -505,7 +533,7 @@
       "outputs": [],
       "source": [
         "best_dnn_run\n",
-        "RunDetails(best_dnn_run).show() "
+        "RunDetails(best_dnn_run).show()"
       ]
     },
     {
@@ -540,7 +568,10 @@
       "outputs": [],
       "source": [
         "from azureml.core import Dataset\n",
-        "test_dataset = Dataset.Tabular.from_delimited_files(path = [(datastore, 'beer-dataset/tabular/test.csv')])\n",
+        "\n",
+        "test_dataset = Dataset.Tabular.from_delimited_files(\n",
+        "    path=[(datastore, \"beer-dataset/tabular/test.csv\")]\n",
+        ")\n",
         "# preview the first 3 rows of the dataset\n",
         "test_dataset.take(5).to_pandas_dataframe()"
       ]
@@ -551,7 +582,7 @@
       "metadata": {},
       "outputs": [],
       "source": [
-        "compute_target = ws.compute_targets['beer-cluster']\n",
+        "compute_target = ws.compute_targets[\"beer-cluster\"]\n",
         "test_experiment = Experiment(ws, experiment_name + \"_test\")"
       ]
     },
@@ -567,9 +598,9 @@
         "import os\n",
         "import shutil\n",
         "\n",
-        "script_folder = os.path.join(os.getcwd(), 'inference')\n",
+        "script_folder = os.path.join(os.getcwd(), \"inference\")\n",
         "os.makedirs(script_folder, exist_ok=True)\n",
-        "shutil.copy('infer.py', script_folder)"
+        "shutil.copy(\"infer.py\", script_folder)"
       ]
     },
     {
@@ -580,8 +611,18 @@
       "source": [
         "from helper import run_inference\n",
         "\n",
-        "test_run = run_inference(test_experiment, compute_target, script_folder, best_dnn_run, test_dataset, valid_dataset, forecast_horizon,\n",
-        "                 target_column_name, time_column_name, freq)"
+        "test_run = run_inference(\n",
+        "    test_experiment,\n",
+        "    compute_target,\n",
+        "    script_folder,\n",
+        "    best_dnn_run,\n",
+        "    test_dataset,\n",
+        "    valid_dataset,\n",
+        "    forecast_horizon,\n",
+        "    target_column_name,\n",
+        "    time_column_name,\n",
+        "    freq,\n",
+        ")"
       ]
     },
     {
@@ -601,8 +642,19 @@
       "source": [
         "from helper import run_multiple_inferences\n",
         "\n",
-        "summary_df = run_multiple_inferences(summary_df, experiment, test_experiment, compute_target, script_folder, test_dataset, \n",
-        "                  valid_dataset, forecast_horizon, target_column_name, time_column_name, freq)"
+        "summary_df = run_multiple_inferences(\n",
+        "    summary_df,\n",
+        "    experiment,\n",
+        "    test_experiment,\n",
+        "    compute_target,\n",
+        "    script_folder,\n",
+        "    test_dataset,\n",
+        "    valid_dataset,\n",
+        "    forecast_horizon,\n",
+        "    target_column_name,\n",
+        "    time_column_name,\n",
+        "    freq,\n",
+        ")"
       ]
     },
     {
@@ -622,7 +674,7 @@
         "    test_run = Run(test_experiment, test_run_id)\n",
         "    test_run.wait_for_completion()\n",
         "    test_score = test_run.get_metrics()[run_summary.primary_metric]\n",
-        "    summary_df.loc[summary_df.run_id == run_id, 'Test Score'] = test_score\n",
+        "    summary_df.loc[summary_df.run_id == run_id, \"Test Score\"] = test_score\n",
         "    print(\"Test Score: \", test_score)"
       ]
     },
@@ -668,7 +720,7 @@
       "name": "python",
       "nbconvert_exporter": "python",
       "pygments_lexer": "ipython3",
-      "version": "3.6.7"
+      "version": "3.6.9"
     }
   },
   "nbformat": 4,

how-to-use-azureml/automated-machine-learning/forecasting-beer-remote/auto-ml-forecasting-beer-remote.yml

@@ -0,0 +1,4 @@
+name: auto-ml-forecasting-beer-remote
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/forecasting-beer-remote/helper.py

@@ -6,120 +6,158 @@ from azureml.core.run import Run
 from azureml.automl.core.shared import constants
 
 
-def split_fraction_by_grain(df, fraction, time_column_name,
-                            grain_column_names=None):
+def split_fraction_by_grain(df, fraction, time_column_name, grain_column_names=None):
     if not grain_column_names:
-        df['tmp_grain_column'] = 'grain'
-        grain_column_names = ['tmp_grain_column']
+        df["tmp_grain_column"] = "grain"
+        grain_column_names = ["tmp_grain_column"]
 
     """Group df by grain and split on last n rows for each group."""
-    df_grouped = (df.sort_values(time_column_name)
-                  .groupby(grain_column_names, group_keys=False))
+    df_grouped = df.sort_values(time_column_name).groupby(
+        grain_column_names, group_keys=False
+    )
 
-    df_head = df_grouped.apply(lambda dfg: dfg.iloc[:-int(len(dfg) *
-                                                          fraction)] if fraction > 0 else dfg)
+    df_head = df_grouped.apply(
+        lambda dfg: dfg.iloc[: -int(len(dfg) * fraction)] if fraction > 0 else dfg
+    )
 
-    df_tail = df_grouped.apply(lambda dfg: dfg.iloc[-int(len(dfg) *
-                                                         fraction):] if fraction > 0 else dfg[:0])
+    df_tail = df_grouped.apply(
+        lambda dfg: dfg.iloc[-int(len(dfg) * fraction) :] if fraction > 0 else dfg[:0]
+    )
 
-    if 'tmp_grain_column' in grain_column_names:
+    if "tmp_grain_column" in grain_column_names:
         for df2 in (df, df_head, df_tail):
-            df2.drop('tmp_grain_column', axis=1, inplace=True)
+            df2.drop("tmp_grain_column", axis=1, inplace=True)
 
-        grain_column_names.remove('tmp_grain_column')
+        grain_column_names.remove("tmp_grain_column")
 
     return df_head, df_tail
 
 
-def split_full_for_forecasting(df, time_column_name,
-                               grain_column_names=None, test_split=0.2):
+def split_full_for_forecasting(
+    df, time_column_name, grain_column_names=None, test_split=0.2
+):
     index_name = df.index.name
 
     # Assumes that there isn't already a column called tmpindex
 
-    df['tmpindex'] = df.index
+    df["tmpindex"] = df.index
 
     train_df, test_df = split_fraction_by_grain(
-        df, test_split, time_column_name, grain_column_names)
+        df, test_split, time_column_name, grain_column_names
+    )
 
-    train_df = train_df.set_index('tmpindex')
+    train_df = train_df.set_index("tmpindex")
     train_df.index.name = index_name
 
-    test_df = test_df.set_index('tmpindex')
+    test_df = test_df.set_index("tmpindex")
     test_df.index.name = index_name
 
-    df.drop('tmpindex', axis=1, inplace=True)
+    df.drop("tmpindex", axis=1, inplace=True)
 
     return train_df, test_df
 
 
 def get_result_df(remote_run):
     children = list(remote_run.get_children(recursive=True))
-    summary_df = pd.DataFrame(index=['run_id', 'run_algorithm',
-                                     'primary_metric', 'Score'])
+    summary_df = pd.DataFrame(
+        index=["run_id", "run_algorithm", "primary_metric", "Score"]
+    )
     goal_minimize = False
     for run in children:
-        if run.get_status().lower() == constants.RunState.COMPLETE_RUN \
-                and 'run_algorithm' in run.properties and 'score' in run.properties:
+        if (
+            run.get_status().lower() == constants.RunState.COMPLETE_RUN
+            and "run_algorithm" in run.properties
+            and "score" in run.properties
+        ):
             # We only count in the completed child runs.
-            summary_df[run.id] = [run.id, run.properties['run_algorithm'],
-                                  run.properties['primary_metric'],
-                                  float(run.properties['score'])]
-            if ('goal' in run.properties):
-                goal_minimize = run.properties['goal'].split('_')[-1] == 'min'
+            summary_df[run.id] = [
+                run.id,
+                run.properties["run_algorithm"],
+                run.properties["primary_metric"],
+                float(run.properties["score"]),
+            ]
+            if "goal" in run.properties:
+                goal_minimize = run.properties["goal"].split("_")[-1] == "min"
 
     summary_df = summary_df.T.sort_values(
-        'Score',
-        ascending=goal_minimize).drop_duplicates(['run_algorithm'])
-    summary_df = summary_df.set_index('run_algorithm')
+        "Score", ascending=goal_minimize
+    ).drop_duplicates(["run_algorithm"])
+    summary_df = summary_df.set_index("run_algorithm")
     return summary_df
 
 
-def run_inference(test_experiment, compute_target, script_folder, train_run,
-                  test_dataset, lookback_dataset, max_horizon,
-                  target_column_name, time_column_name, freq):
-    model_base_name = 'model.pkl'
-    if 'model_data_location' in train_run.properties:
-        model_location = train_run.properties['model_data_location']
-        _, model_base_name = model_location.rsplit('/', 1)
-    train_run.download_file('outputs/{}'.format(model_base_name), 'inference/{}'.format(model_base_name))
-    train_run.download_file('outputs/conda_env_v_1_0_0.yml', 'inference/condafile.yml')
+def run_inference(
+    test_experiment,
+    compute_target,
+    script_folder,
+    train_run,
+    test_dataset,
+    lookback_dataset,
+    max_horizon,
+    target_column_name,
+    time_column_name,
+    freq,
+):
+    model_base_name = "model.pkl"
+    if "model_data_location" in train_run.properties:
+        model_location = train_run.properties["model_data_location"]
+        _, model_base_name = model_location.rsplit("/", 1)
+    train_run.download_file(
+        "outputs/{}".format(model_base_name), "inference/{}".format(model_base_name)
+    )
+    train_run.download_file("outputs/conda_env_v_1_0_0.yml", "inference/condafile.yml")
 
     inference_env = Environment("myenv")
     inference_env.docker.enabled = True
     inference_env.python.conda_dependencies = CondaDependencies(
-        conda_dependencies_file_path='inference/condafile.yml')
+        conda_dependencies_file_path="inference/condafile.yml"
+    )
 
-    est = Estimator(source_directory=script_folder,
-                    entry_script='infer.py',
-                    script_params={
-                        '--max_horizon': max_horizon,
-                        '--target_column_name': target_column_name,
-                        '--time_column_name': time_column_name,
-                        '--frequency': freq,
-                        '--model_path': model_base_name
-                    },
-                    inputs=[test_dataset.as_named_input('test_data'),
-                            lookback_dataset.as_named_input('lookback_data')],
-                    compute_target=compute_target,
-                    environment_definition=inference_env)
+    est = Estimator(
+        source_directory=script_folder,
+        entry_script="infer.py",
+        script_params={
+            "--max_horizon": max_horizon,
+            "--target_column_name": target_column_name,
+            "--time_column_name": time_column_name,
+            "--frequency": freq,
+            "--model_path": model_base_name,
+        },
+        inputs=[
+            test_dataset.as_named_input("test_data"),
+            lookback_dataset.as_named_input("lookback_data"),
+        ],
+        compute_target=compute_target,
+        environment_definition=inference_env,
+    )
 
     run = test_experiment.submit(
-        est, tags={
-            'training_run_id': train_run.id,
-            'run_algorithm': train_run.properties['run_algorithm'],
-            'valid_score': train_run.properties['score'],
-            'primary_metric': train_run.properties['primary_metric']
-        })
+        est,
+        tags={
+            "training_run_id": train_run.id,
+            "run_algorithm": train_run.properties["run_algorithm"],
+            "valid_score": train_run.properties["score"],
+            "primary_metric": train_run.properties["primary_metric"],
+        },
+    )
 
-    run.log("run_algorithm", run.tags['run_algorithm'])
+    run.log("run_algorithm", run.tags["run_algorithm"])
     return run
 
 
-def run_multiple_inferences(summary_df, train_experiment, test_experiment,
-                            compute_target, script_folder, test_dataset,
-                            lookback_dataset, max_horizon, target_column_name,
-                            time_column_name, freq):
+def run_multiple_inferences(
+    summary_df,
+    train_experiment,
+    test_experiment,
+    compute_target,
+    script_folder,
+    test_dataset,
+    lookback_dataset,
+    max_horizon,
+    target_column_name,
+    time_column_name,
+    freq,
+):
     for run_name, run_summary in summary_df.iterrows():
         print(run_name)
         print(run_summary)
@@ -127,12 +165,19 @@ def run_multiple_inferences(summary_df, train_experiment, test_experiment,
         train_run = Run(train_experiment, run_id)
 
         test_run = run_inference(
-            test_experiment, compute_target, script_folder, train_run,
-            test_dataset, lookback_dataset, max_horizon, target_column_name,
-            time_column_name, freq)
+            test_experiment,
+            compute_target,
+            script_folder,
+            train_run,
+            test_dataset,
+            lookback_dataset,
+            max_horizon,
+            target_column_name,
+            time_column_name,
+            freq,
+        )
 
         print(test_run)
-        summary_df.loc[summary_df.run_id == run_id,
-                       'test_run_id'] = test_run.id
+        summary_df.loc[summary_df.run_id == run_id, "test_run_id"] = test_run.id
 
     return summary_df

how-to-use-azureml/automated-machine-learning/forecasting-beer-remote/infer.py

@@ -19,9 +19,14 @@ except ImportError:
     _torch_present = False
 
 
-def align_outputs(y_predicted, X_trans, X_test, y_test,
-                  predicted_column_name='predicted',
-                  horizon_colname='horizon_origin'):
+def align_outputs(
+    y_predicted,
+    X_trans,
+    X_test,
+    y_test,
+    predicted_column_name="predicted",
+    horizon_colname="horizon_origin",
+):
     """
     Demonstrates how to get the output aligned to the inputs
     using pandas indexes. Helps understand what happened if
@@ -33,9 +38,13 @@ def align_outputs(y_predicted, X_trans, X_test, y_test,
     * model was asked to predict past max_horizon -> increase max horizon
     * data at start of X_test was needed for lags -> provide previous periods
     """
-    if (horizon_colname in X_trans):
-        df_fcst = pd.DataFrame({predicted_column_name: y_predicted,
-                                horizon_colname: X_trans[horizon_colname]})
+    if horizon_colname in X_trans:
+        df_fcst = pd.DataFrame(
+            {
+                predicted_column_name: y_predicted,
+                horizon_colname: X_trans[horizon_colname],
+            }
+        )
     else:
         df_fcst = pd.DataFrame({predicted_column_name: y_predicted})
 
@@ -48,20 +57,21 @@ def align_outputs(y_predicted, X_trans, X_test, y_test,
 
     # X_test_full's index does not include origin, so reset for merge
     df_fcst.reset_index(inplace=True)
-    X_test_full = X_test_full.reset_index().drop(columns='index')
-    together = df_fcst.merge(X_test_full, how='right')
+    X_test_full = X_test_full.reset_index().drop(columns="index")
+    together = df_fcst.merge(X_test_full, how="right")
 
     # drop rows where prediction or actuals are nan
     # happens because of missing actuals
     # or at edges of time due to lags/rolling windows
-    clean = together[together[[target_column_name,
-                               predicted_column_name]].notnull().all(axis=1)]
-    return (clean)
+    clean = together[
+        together[[target_column_name, predicted_column_name]].notnull().all(axis=1)
+    ]
+    return clean
 
 
-def do_rolling_forecast_with_lookback(fitted_model, X_test, y_test,
-                                      max_horizon, X_lookback, y_lookback,
-                                      freq='D'):
+def do_rolling_forecast_with_lookback(
+    fitted_model, X_test, y_test, max_horizon, X_lookback, y_lookback, freq="D"
+):
     """
     Produce forecasts on a rolling origin over the given test set.
 
@@ -72,7 +82,7 @@ def do_rolling_forecast_with_lookback(fitted_model, X_test, y_test,
     origin time for constructing lag features.
 
     This function returns a concatenated DataFrame of rolling forecasts.
-     """
+    """
     print("Using lookback of size: ", y_lookback.size)
     df_list = []
     origin_time = X_test[time_column_name].min()
@@ -83,22 +93,28 @@ def do_rolling_forecast_with_lookback(fitted_model, X_test, y_test,
         horizon_time = origin_time + max_horizon * to_offset(freq)
 
         # Extract test data from an expanding window up-to the horizon
-        expand_wind = (X[time_column_name] < horizon_time)
+        expand_wind = X[time_column_name] < horizon_time
         X_test_expand = X[expand_wind]
         y_query_expand = np.zeros(len(X_test_expand)).astype(np.float)
         y_query_expand.fill(np.NaN)
 
         if origin_time != X[time_column_name].min():
             # Set the context by including actuals up-to the origin time
-            test_context_expand_wind = (X[time_column_name] < origin_time)
-            context_expand_wind = (X_test_expand[time_column_name] < origin_time)
+            test_context_expand_wind = X[time_column_name] < origin_time
+            context_expand_wind = X_test_expand[time_column_name] < origin_time
             y_query_expand[context_expand_wind] = y[test_context_expand_wind]
 
         # Print some debug info
-        print("Horizon_time:", horizon_time,
-              " origin_time: ", origin_time,
-              " max_horizon: ", max_horizon,
-              " freq: ", freq)
+        print(
+            "Horizon_time:",
+            horizon_time,
+            " origin_time: ",
+            origin_time,
+            " max_horizon: ",
+            max_horizon,
+            " freq: ",
+            freq,
+        )
         print("expand_wind: ", expand_wind)
         print("y_query_expand")
         print(y_query_expand)
@@ -124,9 +140,14 @@ def do_rolling_forecast_with_lookback(fitted_model, X_test, y_test,
         trans_tindex = X_trans.index.get_level_values(time_column_name)
         trans_roll_wind = (trans_tindex >= origin_time) & (trans_tindex < horizon_time)
         test_roll_wind = expand_wind & (X[time_column_name] >= origin_time)
-        df_list.append(align_outputs(
-            y_fcst[trans_roll_wind], X_trans[trans_roll_wind],
-            X[test_roll_wind], y[test_roll_wind]))
+        df_list.append(
+            align_outputs(
+                y_fcst[trans_roll_wind],
+                X_trans[trans_roll_wind],
+                X[test_roll_wind],
+                y[test_roll_wind],
+            )
+        )
 
         # Advance the origin time
         origin_time = horizon_time
@@ -134,7 +155,7 @@ def do_rolling_forecast_with_lookback(fitted_model, X_test, y_test,
     return pd.concat(df_list, ignore_index=True)
 
 
-def do_rolling_forecast(fitted_model, X_test, y_test, max_horizon, freq='D'):
+def do_rolling_forecast(fitted_model, X_test, y_test, max_horizon, freq="D"):
     """
     Produce forecasts on a rolling origin over the given test set.
 
@@ -145,7 +166,7 @@ def do_rolling_forecast(fitted_model, X_test, y_test, max_horizon, freq='D'):
     origin time for constructing lag features.
 
     This function returns a concatenated DataFrame of rolling forecasts.
-     """
+    """
     df_list = []
     origin_time = X_test[time_column_name].min()
     while origin_time <= X_test[time_column_name].max():
@@ -153,23 +174,28 @@ def do_rolling_forecast(fitted_model, X_test, y_test, max_horizon, freq='D'):
         horizon_time = origin_time + max_horizon * to_offset(freq)
 
         # Extract test data from an expanding window up-to the horizon
-        expand_wind = (X_test[time_column_name] < horizon_time)
+        expand_wind = X_test[time_column_name] < horizon_time
         X_test_expand = X_test[expand_wind]
         y_query_expand = np.zeros(len(X_test_expand)).astype(np.float)
         y_query_expand.fill(np.NaN)
 
         if origin_time != X_test[time_column_name].min():
             # Set the context by including actuals up-to the origin time
-            test_context_expand_wind = (X_test[time_column_name] < origin_time)
-            context_expand_wind = (X_test_expand[time_column_name] < origin_time)
-            y_query_expand[context_expand_wind] = y_test[
-                test_context_expand_wind]
+            test_context_expand_wind = X_test[time_column_name] < origin_time
+            context_expand_wind = X_test_expand[time_column_name] < origin_time
+            y_query_expand[context_expand_wind] = y_test[test_context_expand_wind]
 
         # Print some debug info
-        print("Horizon_time:", horizon_time,
-              " origin_time: ", origin_time,
-              " max_horizon: ", max_horizon,
-              " freq: ", freq)
+        print(
+            "Horizon_time:",
+            horizon_time,
+            " origin_time: ",
+            origin_time,
+            " max_horizon: ",
+            max_horizon,
+            " freq: ",
+            freq,
+        )
         print("expand_wind: ", expand_wind)
         print("y_query_expand")
         print(y_query_expand)
@@ -193,10 +219,14 @@ def do_rolling_forecast(fitted_model, X_test, y_test, max_horizon, freq='D'):
         trans_tindex = X_trans.index.get_level_values(time_column_name)
         trans_roll_wind = (trans_tindex >= origin_time) & (trans_tindex < horizon_time)
         test_roll_wind = expand_wind & (X_test[time_column_name] >= origin_time)
-        df_list.append(align_outputs(y_fcst[trans_roll_wind],
-                                     X_trans[trans_roll_wind],
-                                     X_test[test_roll_wind],
-                                     y_test[test_roll_wind]))
+        df_list.append(
+            align_outputs(
+                y_fcst[trans_roll_wind],
+                X_trans[trans_roll_wind],
+                X_test[test_roll_wind],
+                y_test[test_roll_wind],
+            )
+        )
 
         # Advance the origin time
         origin_time = horizon_time
@@ -230,20 +260,31 @@ def map_location_cuda(storage, loc):
 
 parser = argparse.ArgumentParser()
 parser.add_argument(
-    '--max_horizon', type=int, dest='max_horizon',
-    default=10, help='Max Horizon for forecasting')
+    "--max_horizon",
+    type=int,
+    dest="max_horizon",
+    default=10,
+    help="Max Horizon for forecasting",
+)
 parser.add_argument(
-    '--target_column_name', type=str, dest='target_column_name',
-    help='Target Column Name')
+    "--target_column_name",
+    type=str,
+    dest="target_column_name",
+    help="Target Column Name",
+)
 parser.add_argument(
-    '--time_column_name', type=str, dest='time_column_name',
-    help='Time Column Name')
+    "--time_column_name", type=str, dest="time_column_name", help="Time Column Name"
+)
 parser.add_argument(
-    '--frequency', type=str, dest='freq',
-    help='Frequency of prediction')
+    "--frequency", type=str, dest="freq", help="Frequency of prediction"
+)
 parser.add_argument(
-    '--model_path', type=str, dest='model_path',
-    default='model.pkl', help='Filename of model to be loaded')
+    "--model_path",
+    type=str,
+    dest="model_path",
+    default="model.pkl",
+    help="Filename of model to be loaded",
+)
 
 args = parser.parse_args()
 max_horizon = args.max_horizon
@@ -252,7 +293,7 @@ time_column_name = args.time_column_name
 freq = args.freq
 model_path = args.model_path
 
-print('args passed are: ')
+print("args passed are: ")
 print(max_horizon)
 print(target_column_name)
 print(time_column_name)
@@ -261,39 +302,41 @@ print(model_path)
 
 run = Run.get_context()
 # get input dataset by name
-test_dataset = run.input_datasets['test_data']
-lookback_dataset = run.input_datasets['lookback_data']
+test_dataset = run.input_datasets["test_data"]
+lookback_dataset = run.input_datasets["lookback_data"]
 
 grain_column_names = []
 
 df = test_dataset.to_pandas_dataframe()
 
-print('Read df')
+print("Read df")
 print(df)
 
 X_test_df = test_dataset.drop_columns(columns=[target_column_name])
-y_test_df = test_dataset.with_timestamp_columns(
-    None).keep_columns(columns=[target_column_name])
+y_test_df = test_dataset.with_timestamp_columns(None).keep_columns(
+    columns=[target_column_name]
+)
 
 X_lookback_df = lookback_dataset.drop_columns(columns=[target_column_name])
-y_lookback_df = lookback_dataset.with_timestamp_columns(
-    None).keep_columns(columns=[target_column_name])
+y_lookback_df = lookback_dataset.with_timestamp_columns(None).keep_columns(
+    columns=[target_column_name]
+)
 
 _, ext = os.path.splitext(model_path)
-if ext == '.pt':
+if ext == ".pt":
     # Load the fc-tcn torch model.
     assert _torch_present
     if torch.cuda.is_available():
         map_location = map_location_cuda
     else:
-        map_location = 'cpu'
-    with open(model_path, 'rb') as fh:
+        map_location = "cpu"
+    with open(model_path, "rb") as fh:
         fitted_model = torch.load(fh, map_location=map_location)
 else:
     # Load the sklearn pipeline.
     fitted_model = joblib.load(model_path)
 
-if hasattr(fitted_model, 'get_lookback'):
+if hasattr(fitted_model, "get_lookback"):
     lookback = fitted_model.get_lookback()
     df_all = do_rolling_forecast_with_lookback(
         fitted_model,
@@ -302,26 +345,28 @@ if hasattr(fitted_model, 'get_lookback'):
         max_horizon,
         X_lookback_df.to_pandas_dataframe()[-lookback:],
         y_lookback_df.to_pandas_dataframe().values.T[0][-lookback:],
-        freq)
+        freq,
+    )
 else:
     df_all = do_rolling_forecast(
         fitted_model,
         X_test_df.to_pandas_dataframe(),
         y_test_df.to_pandas_dataframe().values.T[0],
         max_horizon,
-        freq)
+        freq,
+    )
 
 print(df_all)
 
 print("target values:::")
 print(df_all[target_column_name])
 print("predicted values:::")
-print(df_all['predicted'])
+print(df_all["predicted"])
 
 # Use the AutoML scoring module
 regression_metrics = list(constants.REGRESSION_SCALAR_SET)
 y_test = np.array(df_all[target_column_name])
-y_pred = np.array(df_all['predicted'])
+y_pred = np.array(df_all["predicted"])
 scores = scoring.score_regression(y_test, y_pred, regression_metrics)
 
 print("scores:")
@@ -331,12 +376,11 @@ for key, value in scores.items():
     run.log(key, value)
 
 print("Simple forecasting model")
-rmse = np.sqrt(mean_squared_error(
-    df_all[target_column_name], df_all['predicted']))
+rmse = np.sqrt(mean_squared_error(df_all[target_column_name], df_all["predicted"]))
 print("[Test Data] \nRoot Mean squared error: %.2f" % rmse)
-mae = mean_absolute_error(df_all[target_column_name], df_all['predicted'])
-print('mean_absolute_error score: %.2f' % mae)
-print('MAPE: %.2f' % MAPE(df_all[target_column_name], df_all['predicted']))
+mae = mean_absolute_error(df_all[target_column_name], df_all["predicted"])
+print("mean_absolute_error score: %.2f" % mae)
+print("MAPE: %.2f" % MAPE(df_all[target_column_name], df_all["predicted"]))
 
-run.log('rmse', rmse)
-run.log('mae', mae)
+run.log("rmse", rmse)
+run.log("mae", mae)

how-to-use-azureml/automated-machine-learning/forecasting-bike-share/auto-ml-forecasting-bike-share.yml

@@ -0,0 +1,4 @@
+name: auto-ml-forecasting-bike-share
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/forecasting-bike-share/forecasting_script.py

@@ -4,11 +4,14 @@ from sklearn.externals import joblib
 
 parser = argparse.ArgumentParser()
 parser.add_argument(
-    '--target_column_name', type=str, dest='target_column_name',
-    help='Target Column Name')
+    "--target_column_name",
+    type=str,
+    dest="target_column_name",
+    help="Target Column Name",
+)
 parser.add_argument(
-    '--test_dataset', type=str, dest='test_dataset',
-    help='Test Dataset')
+    "--test_dataset", type=str, dest="test_dataset", help="Test Dataset"
+)
 
 args = parser.parse_args()
 target_column_name = args.target_column_name
@@ -20,19 +23,30 @@ ws = run.experiment.workspace
 # get the input dataset by id
 test_dataset = Dataset.get_by_id(ws, id=test_dataset_id)
 
-X_test_df = test_dataset.drop_columns(columns=[target_column_name]).to_pandas_dataframe().reset_index(drop=True)
-y_test_df = test_dataset.with_timestamp_columns(None).keep_columns(columns=[target_column_name]).to_pandas_dataframe()
+X_test_df = (
+    test_dataset.drop_columns(columns=[target_column_name])
+    .to_pandas_dataframe()
+    .reset_index(drop=True)
+)
+y_test_df = (
+    test_dataset.with_timestamp_columns(None)
+    .keep_columns(columns=[target_column_name])
+    .to_pandas_dataframe()
+)
 
-fitted_model = joblib.load('model.pkl')
+fitted_model = joblib.load("model.pkl")
 
 y_pred, X_trans = fitted_model.rolling_evaluation(X_test_df, y_test_df.values)
 
 # Add predictions, actuals, and horizon relative to rolling origin to the test feature data
-assign_dict = {'horizon_origin': X_trans['horizon_origin'].values, 'predicted': y_pred,
-               target_column_name: y_test_df[target_column_name].values}
+assign_dict = {
+    "horizon_origin": X_trans["horizon_origin"].values,
+    "predicted": y_pred,
+    target_column_name: y_test_df[target_column_name].values,
+}
 df_all = X_test_df.assign(**assign_dict)
 
-file_name = 'outputs/predictions.csv'
+file_name = "outputs/predictions.csv"
 export_csv = df_all.to_csv(file_name, header=True)
 
 # Upload the predictions into artifacts

how-to-use-azureml/automated-machine-learning/forecasting-bike-share/run_forecast.py

@@ -1,32 +1,40 @@
 from azureml.core import ScriptRunConfig
 
 
-def run_rolling_forecast(test_experiment, compute_target, train_run,
-                         test_dataset, target_column_name,
-                         inference_folder='./forecast'):
-    train_run.download_file('outputs/model.pkl',
-                            inference_folder + '/model.pkl')
+def run_rolling_forecast(
+    test_experiment,
+    compute_target,
+    train_run,
+    test_dataset,
+    target_column_name,
+    inference_folder="./forecast",
+):
+    train_run.download_file("outputs/model.pkl", inference_folder + "/model.pkl")
 
     inference_env = train_run.get_environment()
 
-    config = ScriptRunConfig(source_directory=inference_folder,
-                             script='forecasting_script.py',
-                             arguments=['--target_column_name',
-                                        target_column_name,
-                                        '--test_dataset',
-                                        test_dataset.as_named_input(test_dataset.name)],
-                             compute_target=compute_target,
-                             environment=inference_env)
+    config = ScriptRunConfig(
+        source_directory=inference_folder,
+        script="forecasting_script.py",
+        arguments=[
+            "--target_column_name",
+            target_column_name,
+            "--test_dataset",
+            test_dataset.as_named_input(test_dataset.name),
+        ],
+        compute_target=compute_target,
+        environment=inference_env,
+    )
 
-    run = test_experiment.submit(config,
-                                 tags={'training_run_id':
-                                       train_run.id,
-                                       'run_algorithm':
-                                       train_run.properties['run_algorithm'],
-                                       'valid_score':
-                                       train_run.properties['score'],
-                                       'primary_metric':
-                                       train_run.properties['primary_metric']})
+    run = test_experiment.submit(
+        config,
+        tags={
+            "training_run_id": train_run.id,
+            "run_algorithm": train_run.properties["run_algorithm"],
+            "valid_score": train_run.properties["score"],
+            "primary_metric": train_run.properties["primary_metric"],
+        },
+    )
 
-    run.log("run_algorithm", run.tags['run_algorithm'])
+    run.log("run_algorithm", run.tags["run_algorithm"])
     return run

how-to-use-azureml/automated-machine-learning/forecasting-energy-demand/auto-ml-forecasting-energy-demand.yml

@@ -0,0 +1,4 @@
+name: auto-ml-forecasting-energy-demand
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/forecasting-energy-demand/forecasting_helper.py

@@ -1,44 +0,0 @@
-import pandas as pd
-import numpy as np
-from pandas.tseries.frequencies import to_offset
-
-
-def align_outputs(y_predicted, X_trans, X_test, y_test, target_column_name,
-                  predicted_column_name='predicted',
-                  horizon_colname='horizon_origin'):
-    """
-    Demonstrates how to get the output aligned to the inputs
-    using pandas indexes. Helps understand what happened if
-    the output's shape differs from the input shape, or if
-    the data got re-sorted by time and grain during forecasting.
-
-    Typical causes of misalignment are:
-    * we predicted some periods that were missing in actuals -> drop from eval
-    * model was asked to predict past max_horizon -> increase max horizon
-    * data at start of X_test was needed for lags -> provide previous periods
-    """
-
-    if (horizon_colname in X_trans):
-        df_fcst = pd.DataFrame({predicted_column_name: y_predicted,
-                                horizon_colname: X_trans[horizon_colname]})
-    else:
-        df_fcst = pd.DataFrame({predicted_column_name: y_predicted})
-
-    # y and X outputs are aligned by forecast() function contract
-    df_fcst.index = X_trans.index
-
-    # align original X_test to y_test
-    X_test_full = X_test.copy()
-    X_test_full[target_column_name] = y_test
-
-    # X_test_full's index does not include origin, so reset for merge
-    df_fcst.reset_index(inplace=True)
-    X_test_full = X_test_full.reset_index().drop(columns='index')
-    together = df_fcst.merge(X_test_full, how='right')
-
-    # drop rows where prediction or actuals are nan
-    # happens because of missing actuals
-    # or at edges of time due to lags/rolling windows
-    clean = together[together[[target_column_name,
-                               predicted_column_name]].notnull().all(axis=1)]
-    return(clean)

how-to-use-azureml/automated-machine-learning/forecasting-energy-demand/forecasting_script.py

@@ -0,0 +1,61 @@
+"""
+This is the script that is executed on the compute instance. It relies
+on the model.pkl file which is uploaded along with this script to the
+compute instance.
+"""
+
+import argparse
+from azureml.core import Dataset, Run
+from sklearn.externals import joblib
+from pandas.tseries.frequencies import to_offset
+
+parser = argparse.ArgumentParser()
+parser.add_argument(
+    "--target_column_name",
+    type=str,
+    dest="target_column_name",
+    help="Target Column Name",
+)
+parser.add_argument(
+    "--test_dataset", type=str, dest="test_dataset", help="Test Dataset"
+)
+
+args = parser.parse_args()
+target_column_name = args.target_column_name
+test_dataset_id = args.test_dataset
+
+run = Run.get_context()
+ws = run.experiment.workspace
+
+# get the input dataset by id
+test_dataset = Dataset.get_by_id(ws, id=test_dataset_id)
+
+X_test = test_dataset.to_pandas_dataframe().reset_index(drop=True)
+y_test = X_test.pop(target_column_name).values
+
+# generate forecast
+fitted_model = joblib.load("model.pkl")
+# We have default quantiles values set as below(95th percentile)
+quantiles = [0.025, 0.5, 0.975]
+predicted_column_name = "predicted"
+PI = "prediction_interval"
+fitted_model.quantiles = quantiles
+pred_quantiles = fitted_model.forecast_quantiles(X_test)
+pred_quantiles[PI] = pred_quantiles[[min(quantiles), max(quantiles)]].apply(
+    lambda x: "[{}, {}]".format(x[0], x[1]), axis=1
+)
+X_test[target_column_name] = y_test
+X_test[PI] = pred_quantiles[PI]
+X_test[predicted_column_name] = pred_quantiles[0.5]
+# drop rows where prediction or actuals are nan
+# happens because of missing actuals
+# or at edges of time due to lags/rolling windows
+clean = X_test[
+    X_test[[target_column_name, predicted_column_name]].notnull().all(axis=1)
+]
+
+file_name = "outputs/predictions.csv"
+export_csv = clean.to_csv(file_name, header=True, index=False)  # added Index
+
+# Upload the predictions into artifacts
+run.upload_file(name=file_name, path_or_stream=file_name)

how-to-use-azureml/automated-machine-learning/forecasting-energy-demand/metrics_helper.py

@@ -1,22 +0,0 @@
-import pandas as pd
-import numpy as np
-
-
-def APE(actual, pred):
-    """
-    Calculate absolute percentage error.
-    Returns a vector of APE values with same length as actual/pred.
-    """
-    return 100 * np.abs((actual - pred) / actual)
-
-
-def MAPE(actual, pred):
-    """
-    Calculate mean absolute percentage error.
-    Remove NA and values where actual is close to zero
-    """
-    not_na = ~(np.isnan(actual) | np.isnan(pred))
-    not_zero = ~np.isclose(actual, 0.0)
-    actual_safe = actual[not_na & not_zero]
-    pred_safe = pred[not_na & not_zero]
-    return np.mean(APE(actual_safe, pred_safe))

how-to-use-azureml/automated-machine-learning/forecasting-energy-demand/run_forecast.py

@@ -0,0 +1,49 @@
+import os
+import shutil
+from azureml.core import ScriptRunConfig
+
+
+def run_remote_inference(
+    test_experiment,
+    compute_target,
+    train_run,
+    test_dataset,
+    target_column_name,
+    inference_folder="./forecast",
+):
+    # Create local directory to copy the model.pkl and forecsting_script.py files into.
+    # These files will be uploaded to and executed on the compute instance.
+    os.makedirs(inference_folder, exist_ok=True)
+    shutil.copy("forecasting_script.py", inference_folder)
+
+    train_run.download_file(
+        "outputs/model.pkl", os.path.join(inference_folder, "model.pkl")
+    )
+
+    inference_env = train_run.get_environment()
+
+    config = ScriptRunConfig(
+        source_directory=inference_folder,
+        script="forecasting_script.py",
+        arguments=[
+            "--target_column_name",
+            target_column_name,
+            "--test_dataset",
+            test_dataset.as_named_input(test_dataset.name),
+        ],
+        compute_target=compute_target,
+        environment=inference_env,
+    )
+
+    run = test_experiment.submit(
+        config,
+        tags={
+            "training_run_id": train_run.id,
+            "run_algorithm": train_run.properties["run_algorithm"],
+            "valid_score": train_run.properties["score"],
+            "primary_metric": train_run.properties["primary_metric"],
+        },
+    )
+
+    run.log("run_algorithm", run.tags["run_algorithm"])
+    return run

how-to-use-azureml/automated-machine-learning/forecasting-forecast-function/auto-ml-forecasting-function.yml

@@ -0,0 +1,4 @@
+name: auto-ml-forecasting-function
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/forecasting-github-dau/auto-ml-forecasting-github-dau.ipynb

@@ -0,0 +1,725 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "Copyright (c) Microsoft Corporation. All rights reserved.\n",
+    "\n",
+    "Licensed under the MIT License."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "![Impressions](https://PixelServer20190423114238.azurewebsites.net/api/impressions/MachineLearningNotebooks/how-to-use-azureml/automated-machine-learning/forecasting-beer-remote/auto-ml-forecasting-beer-remote.png)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "# Automated Machine Learning\n",
+    "**Github DAU Forecasting**\n",
+    "\n",
+    "## Contents\n",
+    "1. [Introduction](#Introduction)\n",
+    "1. [Setup](#Setup)\n",
+    "1. [Data](#Data)\n",
+    "1. [Train](#Train)\n",
+    "1. [Evaluate](#Evaluate)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "## Introduction\n",
+    "This notebook demonstrates demand forecasting for Github Daily Active Users Dataset using AutoML.\n",
+    "\n",
+    "AutoML highlights here include using Deep Learning forecasts, Arima, Prophet,  Remote Execution and Remote Inferencing, and working with the `forecast` function. Please also look at the additional forecasting notebooks, which document lagging, rolling windows, forecast quantiles, other ways to use the forecast function, and forecaster deployment.\n",
+    "\n",
+    "Make sure you have executed the [configuration](../../../configuration.ipynb) before running this notebook.\n",
+    "\n",
+    "Notebook synopsis:\n",
+    "\n",
+    "1. Creating an Experiment in an existing Workspace\n",
+    "2. Configuration and remote run of AutoML for a time-series model exploring Regression learners, Arima, Prophet and DNNs\n",
+    "4. Evaluating the fitted model using a rolling test "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "## Setup\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "import azureml.core\n",
+    "import pandas as pd\n",
+    "import numpy as np\n",
+    "import logging\n",
+    "import warnings\n",
+    "\n",
+    "from pandas.tseries.frequencies import to_offset\n",
+    "\n",
+    "# Squash warning messages for cleaner output in the notebook\n",
+    "warnings.showwarning = lambda *args, **kwargs: None\n",
+    "\n",
+    "from azureml.core.workspace import Workspace\n",
+    "from azureml.core.experiment import Experiment\n",
+    "from azureml.train.automl import AutoMLConfig\n",
+    "from matplotlib import pyplot as plt\n",
+    "from sklearn.metrics import mean_absolute_error, mean_squared_error\n",
+    "from azureml.train.estimator import Estimator"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "This notebook is compatible with Azure ML SDK version 1.35.0 or later."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "print(\"You are currently using version\", azureml.core.VERSION, \"of the Azure ML SDK\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "As part of the setup you have already created a <b>Workspace</b>. To run AutoML, you also need to create an <b>Experiment</b>. An Experiment corresponds to a prediction problem you are trying to solve, while a Run corresponds to a specific approach to the problem."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "ws = Workspace.from_config()\n",
+    "\n",
+    "# choose a name for the run history container in the workspace\n",
+    "experiment_name = \"github-remote-cpu\"\n",
+    "\n",
+    "experiment = Experiment(ws, experiment_name)\n",
+    "\n",
+    "output = {}\n",
+    "output[\"Subscription ID\"] = ws.subscription_id\n",
+    "output[\"Workspace\"] = ws.name\n",
+    "output[\"Resource Group\"] = ws.resource_group\n",
+    "output[\"Location\"] = ws.location\n",
+    "output[\"Run History Name\"] = experiment_name\n",
+    "pd.set_option(\"display.max_colwidth\", -1)\n",
+    "outputDf = pd.DataFrame(data=output, index=[\"\"])\n",
+    "outputDf.T"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "### Using AmlCompute\n",
+    "You will need to create a [compute target](https://docs.microsoft.com/azure/machine-learning/service/concept-azure-machine-learning-architecture#compute-target) for your AutoML run. In this tutorial, you use `AmlCompute` as your training compute resource.\n",
+    "\n",
+    "> Note that if you have an AzureML Data Scientist role, you will not have permission to create compute resources. Talk to your workspace or IT admin to create the compute targets described in this section, if they do not already exist."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.core.compute import ComputeTarget, AmlCompute\n",
+    "from azureml.core.compute_target import ComputeTargetException\n",
+    "\n",
+    "# Choose a name for your CPU cluster\n",
+    "cpu_cluster_name = \"github-cluster\"\n",
+    "\n",
+    "# Verify that cluster does not exist already\n",
+    "try:\n",
+    "    compute_target = ComputeTarget(workspace=ws, name=cpu_cluster_name)\n",
+    "    print(\"Found existing cluster, use it.\")\n",
+    "except ComputeTargetException:\n",
+    "    compute_config = AmlCompute.provisioning_configuration(\n",
+    "        vm_size=\"STANDARD_DS12_V2\", max_nodes=4\n",
+    "    )\n",
+    "    compute_target = ComputeTarget.create(ws, cpu_cluster_name, compute_config)\n",
+    "\n",
+    "compute_target.wait_for_completion(show_output=True)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "## Data\n",
+    "Read Github DAU data from file, and preview data."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "Let's set up what we know about the dataset. \n",
+    "\n",
+    "**Target column** is what we want to forecast.\n",
+    "\n",
+    "**Time column** is the time axis along which to predict.\n",
+    "\n",
+    "**Time series identifier columns** are identified by values of the columns listed `time_series_id_column_names`, for example \"store\" and \"item\" if your data has multiple time series of sales, one series for each combination of store and item sold.\n",
+    "\n",
+    "**Forecast frequency (freq)** This optional parameter represents the period with which the forecast is desired, for example, daily, weekly, yearly, etc. Use this parameter for the correction of time series containing irregular data points or for padding of short time series. The frequency needs to be a pandas offset alias. Please refer to [pandas documentation](https://pandas.pydata.org/pandas-docs/stable/user_guide/timeseries.html#dateoffset-objects) for more information.\n",
+    "\n",
+    "This dataset has only one time series. Please see the [orange juice notebook](https://github.com/Azure/MachineLearningNotebooks/tree/master/how-to-use-azureml/automated-machine-learning/forecasting-orange-juice-sales) for an example of a multi-time series dataset."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "import pandas as pd\n",
+    "from pandas import DataFrame\n",
+    "from pandas import Grouper\n",
+    "from pandas import concat\n",
+    "from pandas.plotting import register_matplotlib_converters\n",
+    "\n",
+    "register_matplotlib_converters()\n",
+    "plt.figure(figsize=(20, 10))\n",
+    "plt.tight_layout()\n",
+    "\n",
+    "plt.subplot(2, 1, 1)\n",
+    "plt.title(\"Github Daily Active User By Year\")\n",
+    "df = pd.read_csv(\"github_dau_2011-2018_train.csv\", parse_dates=True, index_col=\"date\")\n",
+    "test_df = pd.read_csv(\n",
+    "    \"github_dau_2011-2018_test.csv\", parse_dates=True, index_col=\"date\"\n",
+    ")\n",
+    "plt.plot(df)\n",
+    "\n",
+    "plt.subplot(2, 1, 2)\n",
+    "plt.title(\"Github Daily Active User By Month\")\n",
+    "groups = df.groupby(df.index.month)\n",
+    "months = concat([DataFrame(x[1].values) for x in groups], axis=1)\n",
+    "months = DataFrame(months)\n",
+    "months.columns = range(1, 49)\n",
+    "months.boxplot()\n",
+    "\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "target_column_name = \"count\"\n",
+    "time_column_name = \"date\"\n",
+    "time_series_id_column_names = []\n",
+    "freq = \"D\"  # Daily data"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Split Training data into Train and Validation set and Upload to Datastores"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "from helper import split_fraction_by_grain\n",
+    "from helper import split_full_for_forecasting\n",
+    "\n",
+    "train, valid = split_full_for_forecasting(df, time_column_name)\n",
+    "train.to_csv(\"train.csv\")\n",
+    "valid.to_csv(\"valid.csv\")\n",
+    "test_df.to_csv(\"test.csv\")\n",
+    "\n",
+    "datastore = ws.get_default_datastore()\n",
+    "datastore.upload_files(\n",
+    "    files=[\"./train.csv\"],\n",
+    "    target_path=\"github-dataset/tabular/\",\n",
+    "    overwrite=True,\n",
+    "    show_progress=True,\n",
+    ")\n",
+    "datastore.upload_files(\n",
+    "    files=[\"./valid.csv\"],\n",
+    "    target_path=\"github-dataset/tabular/\",\n",
+    "    overwrite=True,\n",
+    "    show_progress=True,\n",
+    ")\n",
+    "datastore.upload_files(\n",
+    "    files=[\"./test.csv\"],\n",
+    "    target_path=\"github-dataset/tabular/\",\n",
+    "    overwrite=True,\n",
+    "    show_progress=True,\n",
+    ")\n",
+    "\n",
+    "from azureml.core import Dataset\n",
+    "\n",
+    "train_dataset = Dataset.Tabular.from_delimited_files(\n",
+    "    path=[(datastore, \"github-dataset/tabular/train.csv\")]\n",
+    ")\n",
+    "valid_dataset = Dataset.Tabular.from_delimited_files(\n",
+    "    path=[(datastore, \"github-dataset/tabular/valid.csv\")]\n",
+    ")\n",
+    "test_dataset = Dataset.Tabular.from_delimited_files(\n",
+    "    path=[(datastore, \"github-dataset/tabular/test.csv\")]\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "### Setting forecaster maximum horizon \n",
+    "\n",
+    "The forecast horizon is the number of periods into the future that the model should predict. Here, we set the horizon to 12 periods (i.e. 12 months). Notice that this is much shorter than the number of months in the test set; we will need to use a rolling test to evaluate the performance on the whole test set. For more discussion of forecast horizons and guiding principles for setting them, please see the [energy demand notebook](https://github.com/Azure/MachineLearningNotebooks/tree/master/how-to-use-azureml/automated-machine-learning/forecasting-energy-demand).  "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "forecast_horizon = 12"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "## Train\n",
+    "\n",
+    "Instantiate a AutoMLConfig object. This defines the settings and data used to run the experiment.\n",
+    "\n",
+    "|Property|Description|\n",
+    "|-|-|\n",
+    "|**task**|forecasting|\n",
+    "|**primary_metric**|This is the metric that you want to optimize.<br> Forecasting supports the following primary metrics <br><i>spearman_correlation</i><br><i>normalized_root_mean_squared_error</i><br><i>r2_score</i><br><i>normalized_mean_absolute_error</i>\n",
+    "|**iteration_timeout_minutes**|Time limit in minutes for each iteration.|\n",
+    "|**training_data**|Input dataset, containing both features and label column.|\n",
+    "|**label_column_name**|The name of the label column.|\n",
+    "|**enable_dnn**|Enable Forecasting DNNs|\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.automl.core.forecasting_parameters import ForecastingParameters\n",
+    "\n",
+    "forecasting_parameters = ForecastingParameters(\n",
+    "    time_column_name=time_column_name,\n",
+    "    forecast_horizon=forecast_horizon,\n",
+    "    freq=\"D\",  # Set the forecast frequency to be daily\n",
+    ")\n",
+    "\n",
+    "# We will disable the enable_early_stopping flag to ensure the DNN model is recommended for demonstration purpose.\n",
+    "automl_config = AutoMLConfig(\n",
+    "    task=\"forecasting\",\n",
+    "    primary_metric=\"normalized_root_mean_squared_error\",\n",
+    "    experiment_timeout_hours=1,\n",
+    "    training_data=train_dataset,\n",
+    "    label_column_name=target_column_name,\n",
+    "    validation_data=valid_dataset,\n",
+    "    verbosity=logging.INFO,\n",
+    "    compute_target=compute_target,\n",
+    "    max_concurrent_iterations=4,\n",
+    "    max_cores_per_iteration=-1,\n",
+    "    enable_dnn=True,\n",
+    "    enable_early_stopping=False,\n",
+    "    forecasting_parameters=forecasting_parameters,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "We will now run the experiment, starting with 10 iterations of model search. The experiment can be continued for more iterations if more accurate results are required. Validation errors and current status will be shown when setting `show_output=True` and the execution will be synchronous."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "remote_run = experiment.submit(automl_config, show_output=True)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "# If you need to retrieve a run that already started, use the following code\n",
+    "# from azureml.train.automl.run import AutoMLRun\n",
+    "# remote_run = AutoMLRun(experiment = experiment, run_id = '<replace with your run id>')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "Displaying the run objects gives you links to the visual tools in the Azure Portal. Go try them!"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "### Retrieve the Best Model for Each Algorithm\n",
+    "Below we select the best pipeline from our iterations. The get_output method on automl_classifier returns the best run and the fitted model for the last fit invocation. There are overloads on get_output that allow you to retrieve the best run and fitted model for any logged metric or a particular iteration."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "from helper import get_result_df\n",
+    "\n",
+    "summary_df = get_result_df(remote_run)\n",
+    "summary_df"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.core.run import Run\n",
+    "from azureml.widgets import RunDetails\n",
+    "\n",
+    "forecast_model = \"TCNForecaster\"\n",
+    "if not forecast_model in summary_df[\"run_id\"]:\n",
+    "    forecast_model = \"ForecastTCN\"\n",
+    "\n",
+    "best_dnn_run_id = summary_df[\"run_id\"][forecast_model]\n",
+    "best_dnn_run = Run(experiment, best_dnn_run_id)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "best_dnn_run.parent\n",
+    "RunDetails(best_dnn_run.parent).show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "best_dnn_run\n",
+    "RunDetails(best_dnn_run).show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "## Evaluate on Test Data"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "source": [
+    "We now use the best fitted model from the AutoML Run to make forecasts for the test set.  \n",
+    "\n",
+    "We always score on the original dataset whose schema matches the training set schema."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.core import Dataset\n",
+    "\n",
+    "test_dataset = Dataset.Tabular.from_delimited_files(\n",
+    "    path=[(datastore, \"github-dataset/tabular/test.csv\")]\n",
+    ")\n",
+    "# preview the first 3 rows of the dataset\n",
+    "test_dataset.take(5).to_pandas_dataframe()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "compute_target = ws.compute_targets[\"github-cluster\"]\n",
+    "test_experiment = Experiment(ws, experiment_name + \"_test\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "import shutil\n",
+    "\n",
+    "script_folder = os.path.join(os.getcwd(), \"inference\")\n",
+    "os.makedirs(script_folder, exist_ok=True)\n",
+    "shutil.copy(\"infer.py\", script_folder)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from helper import run_inference\n",
+    "\n",
+    "test_run = run_inference(\n",
+    "    test_experiment,\n",
+    "    compute_target,\n",
+    "    script_folder,\n",
+    "    best_dnn_run,\n",
+    "    test_dataset,\n",
+    "    valid_dataset,\n",
+    "    forecast_horizon,\n",
+    "    target_column_name,\n",
+    "    time_column_name,\n",
+    "    freq,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "RunDetails(test_run).show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from helper import run_multiple_inferences\n",
+    "\n",
+    "summary_df = run_multiple_inferences(\n",
+    "    summary_df,\n",
+    "    experiment,\n",
+    "    test_experiment,\n",
+    "    compute_target,\n",
+    "    script_folder,\n",
+    "    test_dataset,\n",
+    "    valid_dataset,\n",
+    "    forecast_horizon,\n",
+    "    target_column_name,\n",
+    "    time_column_name,\n",
+    "    freq,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "for run_name, run_summary in summary_df.iterrows():\n",
+    "    print(run_name)\n",
+    "    print(run_summary)\n",
+    "    run_id = run_summary.run_id\n",
+    "    test_run_id = run_summary.test_run_id\n",
+    "    test_run = Run(test_experiment, test_run_id)\n",
+    "    test_run.wait_for_completion()\n",
+    "    test_score = test_run.get_metrics()[run_summary.primary_metric]\n",
+    "    summary_df.loc[summary_df.run_id == run_id, \"Test Score\"] = test_score\n",
+    "    print(\"Test Score: \", test_score)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "hideCode": false,
+    "hidePrompt": false
+   },
+   "outputs": [],
+   "source": [
+    "summary_df"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "authors": [
+   {
+    "name": "jialiu"
+   }
+  ],
+  "hide_code_all_hidden": false,
+  "kernelspec": {
+   "display_name": "Python 3.6 - AzureML",
+   "language": "python",
+   "name": "python3-azureml"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.6.9"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

how-to-use-azureml/automated-machine-learning/forecasting-github-dau/github_dau_2011-2018_test.csv

@@ -0,0 +1,455 @@
+date,count,day_of_week,month_of_year,holiday
+2017-06-04,104663,6.0,5.0,0.0
+2017-06-05,155824,0.0,5.0,0.0
+2017-06-06,164908,1.0,5.0,0.0
+2017-06-07,170309,2.0,5.0,0.0
+2017-06-08,164256,3.0,5.0,0.0
+2017-06-09,153406,4.0,5.0,0.0
+2017-06-10,97024,5.0,5.0,0.0
+2017-06-11,103442,6.0,5.0,0.0
+2017-06-12,160768,0.0,5.0,0.0
+2017-06-13,166288,1.0,5.0,0.0
+2017-06-14,163819,2.0,5.0,0.0
+2017-06-15,157593,3.0,5.0,0.0
+2017-06-16,149259,4.0,5.0,0.0
+2017-06-17,95579,5.0,5.0,0.0
+2017-06-18,98723,6.0,5.0,0.0
+2017-06-19,159076,0.0,5.0,0.0
+2017-06-20,163340,1.0,5.0,0.0
+2017-06-21,163344,2.0,5.0,0.0
+2017-06-22,159528,3.0,5.0,0.0
+2017-06-23,146563,4.0,5.0,0.0
+2017-06-24,92631,5.0,5.0,0.0
+2017-06-25,96549,6.0,5.0,0.0
+2017-06-26,153249,0.0,5.0,0.0
+2017-06-27,160357,1.0,5.0,0.0
+2017-06-28,159941,2.0,5.0,0.0
+2017-06-29,156781,3.0,5.0,0.0
+2017-06-30,144709,4.0,5.0,0.0
+2017-07-01,89101,5.0,6.0,0.0
+2017-07-02,93046,6.0,6.0,0.0
+2017-07-03,144113,0.0,6.0,0.0
+2017-07-04,143061,1.0,6.0,1.0
+2017-07-05,154603,2.0,6.0,0.0
+2017-07-06,157200,3.0,6.0,0.0
+2017-07-07,147213,4.0,6.0,0.0
+2017-07-08,92348,5.0,6.0,0.0
+2017-07-09,97018,6.0,6.0,0.0
+2017-07-10,157192,0.0,6.0,0.0
+2017-07-11,161819,1.0,6.0,0.0
+2017-07-12,161998,2.0,6.0,0.0
+2017-07-13,160280,3.0,6.0,0.0
+2017-07-14,146818,4.0,6.0,0.0
+2017-07-15,93041,5.0,6.0,0.0
+2017-07-16,97505,6.0,6.0,0.0
+2017-07-17,156167,0.0,6.0,0.0
+2017-07-18,162855,1.0,6.0,0.0
+2017-07-19,162519,2.0,6.0,0.0
+2017-07-20,159941,3.0,6.0,0.0
+2017-07-21,148460,4.0,6.0,0.0
+2017-07-22,93431,5.0,6.0,0.0
+2017-07-23,98553,6.0,6.0,0.0
+2017-07-24,156202,0.0,6.0,0.0
+2017-07-25,162503,1.0,6.0,0.0
+2017-07-26,158479,2.0,6.0,0.0
+2017-07-27,158192,3.0,6.0,0.0
+2017-07-28,147108,4.0,6.0,0.0
+2017-07-29,93799,5.0,6.0,0.0
+2017-07-30,97920,6.0,6.0,0.0
+2017-07-31,152197,0.0,6.0,0.0
+2017-08-01,158477,1.0,7.0,0.0
+2017-08-02,159089,2.0,7.0,0.0
+2017-08-03,157182,3.0,7.0,0.0
+2017-08-04,146345,4.0,7.0,0.0
+2017-08-05,92534,5.0,7.0,0.0
+2017-08-06,97128,6.0,7.0,0.0
+2017-08-07,151359,0.0,7.0,0.0
+2017-08-08,159895,1.0,7.0,0.0
+2017-08-09,158329,2.0,7.0,0.0
+2017-08-10,155468,3.0,7.0,0.0
+2017-08-11,144914,4.0,7.0,0.0
+2017-08-12,92258,5.0,7.0,0.0
+2017-08-13,95933,6.0,7.0,0.0
+2017-08-14,147706,0.0,7.0,0.0
+2017-08-15,151115,1.0,7.0,0.0
+2017-08-16,157640,2.0,7.0,0.0
+2017-08-17,156600,3.0,7.0,0.0
+2017-08-18,146980,4.0,7.0,0.0
+2017-08-19,94592,5.0,7.0,0.0
+2017-08-20,99320,6.0,7.0,0.0
+2017-08-21,145727,0.0,7.0,0.0
+2017-08-22,160260,1.0,7.0,0.0
+2017-08-23,160440,2.0,7.0,0.0
+2017-08-24,157830,3.0,7.0,0.0
+2017-08-25,145822,4.0,7.0,0.0
+2017-08-26,94706,5.0,7.0,0.0
+2017-08-27,99047,6.0,7.0,0.0
+2017-08-28,152112,0.0,7.0,0.0
+2017-08-29,162440,1.0,7.0,0.0
+2017-08-30,162902,2.0,7.0,0.0
+2017-08-31,159498,3.0,7.0,0.0
+2017-09-01,145689,4.0,8.0,0.0
+2017-09-02,93589,5.0,8.0,0.0
+2017-09-03,100058,6.0,8.0,0.0
+2017-09-04,140865,0.0,8.0,1.0
+2017-09-05,165715,1.0,8.0,0.0
+2017-09-06,167463,2.0,8.0,0.0
+2017-09-07,164811,3.0,8.0,0.0
+2017-09-08,156157,4.0,8.0,0.0
+2017-09-09,101358,5.0,8.0,0.0
+2017-09-10,107915,6.0,8.0,0.0
+2017-09-11,167845,0.0,8.0,0.0
+2017-09-12,172756,1.0,8.0,0.0
+2017-09-13,172851,2.0,8.0,0.0
+2017-09-14,171675,3.0,8.0,0.0
+2017-09-15,159266,4.0,8.0,0.0
+2017-09-16,103547,5.0,8.0,0.0
+2017-09-17,110964,6.0,8.0,0.0
+2017-09-18,170976,0.0,8.0,0.0
+2017-09-19,177864,1.0,8.0,0.0
+2017-09-20,173567,2.0,8.0,0.0
+2017-09-21,172017,3.0,8.0,0.0
+2017-09-22,161357,4.0,8.0,0.0
+2017-09-23,104681,5.0,8.0,0.0
+2017-09-24,111711,6.0,8.0,0.0
+2017-09-25,173517,0.0,8.0,0.0
+2017-09-26,180049,1.0,8.0,0.0
+2017-09-27,178307,2.0,8.0,0.0
+2017-09-28,174157,3.0,8.0,0.0
+2017-09-29,161707,4.0,8.0,0.0
+2017-09-30,110536,5.0,8.0,0.0
+2017-10-01,106505,6.0,9.0,0.0
+2017-10-02,157565,0.0,9.0,0.0
+2017-10-03,164764,1.0,9.0,0.0
+2017-10-04,163383,2.0,9.0,0.0
+2017-10-05,162847,3.0,9.0,0.0
+2017-10-06,153575,4.0,9.0,0.0
+2017-10-07,107472,5.0,9.0,0.0
+2017-10-08,116127,6.0,9.0,0.0
+2017-10-09,174457,0.0,9.0,1.0
+2017-10-10,185217,1.0,9.0,0.0
+2017-10-11,185120,2.0,9.0,0.0
+2017-10-12,180844,3.0,9.0,0.0
+2017-10-13,170178,4.0,9.0,0.0
+2017-10-14,112754,5.0,9.0,0.0
+2017-10-15,121251,6.0,9.0,0.0
+2017-10-16,183906,0.0,9.0,0.0
+2017-10-17,188945,1.0,9.0,0.0
+2017-10-18,187297,2.0,9.0,0.0
+2017-10-19,183867,3.0,9.0,0.0
+2017-10-20,173021,4.0,9.0,0.0
+2017-10-21,115851,5.0,9.0,0.0
+2017-10-22,126088,6.0,9.0,0.0
+2017-10-23,189452,0.0,9.0,0.0
+2017-10-24,194412,1.0,9.0,0.0
+2017-10-25,192293,2.0,9.0,0.0
+2017-10-26,190163,3.0,9.0,0.0
+2017-10-27,177053,4.0,9.0,0.0
+2017-10-28,114934,5.0,9.0,0.0
+2017-10-29,125289,6.0,9.0,0.0
+2017-10-30,189245,0.0,9.0,0.0
+2017-10-31,191480,1.0,9.0,0.0
+2017-11-01,182281,2.0,10.0,0.0
+2017-11-02,186351,3.0,10.0,0.0
+2017-11-03,175422,4.0,10.0,0.0
+2017-11-04,118160,5.0,10.0,0.0
+2017-11-05,127602,6.0,10.0,0.0
+2017-11-06,191067,0.0,10.0,0.0
+2017-11-07,197083,1.0,10.0,0.0
+2017-11-08,194333,2.0,10.0,0.0
+2017-11-09,193914,3.0,10.0,0.0
+2017-11-10,179933,4.0,10.0,1.0
+2017-11-11,121346,5.0,10.0,0.0
+2017-11-12,131900,6.0,10.0,0.0
+2017-11-13,196969,0.0,10.0,0.0
+2017-11-14,201949,1.0,10.0,0.0
+2017-11-15,198424,2.0,10.0,0.0
+2017-11-16,196902,3.0,10.0,0.0
+2017-11-17,183893,4.0,10.0,0.0
+2017-11-18,122767,5.0,10.0,0.0
+2017-11-19,130890,6.0,10.0,0.0
+2017-11-20,194515,0.0,10.0,0.0
+2017-11-21,198601,1.0,10.0,0.0
+2017-11-22,191041,2.0,10.0,0.0
+2017-11-23,170321,3.0,10.0,1.0
+2017-11-24,155623,4.0,10.0,0.0
+2017-11-25,115759,5.0,10.0,0.0
+2017-11-26,128771,6.0,10.0,0.0
+2017-11-27,199419,0.0,10.0,0.0
+2017-11-28,207253,1.0,10.0,0.0
+2017-11-29,205406,2.0,10.0,0.0
+2017-11-30,200674,3.0,10.0,0.0
+2017-12-01,187017,4.0,11.0,0.0
+2017-12-02,129735,5.0,11.0,0.0
+2017-12-03,139120,6.0,11.0,0.0
+2017-12-04,205505,0.0,11.0,0.0
+2017-12-05,208218,1.0,11.0,0.0
+2017-12-06,202480,2.0,11.0,0.0
+2017-12-07,197822,3.0,11.0,0.0
+2017-12-08,180686,4.0,11.0,0.0
+2017-12-09,123667,5.0,11.0,0.0
+2017-12-10,130987,6.0,11.0,0.0
+2017-12-11,193901,0.0,11.0,0.0
+2017-12-12,194997,1.0,11.0,0.0
+2017-12-13,192063,2.0,11.0,0.0
+2017-12-14,186496,3.0,11.0,0.0
+2017-12-15,170812,4.0,11.0,0.0
+2017-12-16,110474,5.0,11.0,0.0
+2017-12-17,118165,6.0,11.0,0.0
+2017-12-18,176843,0.0,11.0,0.0
+2017-12-19,179550,1.0,11.0,0.0
+2017-12-20,173506,2.0,11.0,0.0
+2017-12-21,165910,3.0,11.0,0.0
+2017-12-22,145886,4.0,11.0,0.0
+2017-12-23,95246,5.0,11.0,0.0
+2017-12-24,88781,6.0,11.0,0.0
+2017-12-25,98189,0.0,11.0,1.0
+2017-12-26,121383,1.0,11.0,0.0
+2017-12-27,135300,2.0,11.0,0.0
+2017-12-28,136827,3.0,11.0,0.0
+2017-12-29,127700,4.0,11.0,0.0
+2017-12-30,93014,5.0,11.0,0.0
+2017-12-31,82878,6.0,11.0,0.0
+2018-01-01,86419,0.0,0.0,1.0
+2018-01-02,147428,1.0,0.0,0.0
+2018-01-03,162193,2.0,0.0,0.0
+2018-01-04,163784,3.0,0.0,0.0
+2018-01-05,158606,4.0,0.0,0.0
+2018-01-06,113467,5.0,0.0,0.0
+2018-01-07,118313,6.0,0.0,0.0
+2018-01-08,175623,0.0,0.0,0.0
+2018-01-09,183880,1.0,0.0,0.0
+2018-01-10,183945,2.0,0.0,0.0
+2018-01-11,181769,3.0,0.0,0.0
+2018-01-12,170552,4.0,0.0,0.0
+2018-01-13,115707,5.0,0.0,0.0
+2018-01-14,121191,6.0,0.0,0.0
+2018-01-15,176127,0.0,0.0,1.0
+2018-01-16,188032,1.0,0.0,0.0
+2018-01-17,189871,2.0,0.0,0.0
+2018-01-18,189348,3.0,0.0,0.0
+2018-01-19,177456,4.0,0.0,0.0
+2018-01-20,123321,5.0,0.0,0.0
+2018-01-21,128306,6.0,0.0,0.0
+2018-01-22,186132,0.0,0.0,0.0
+2018-01-23,197618,1.0,0.0,0.0
+2018-01-24,196402,2.0,0.0,0.0
+2018-01-25,192722,3.0,0.0,0.0
+2018-01-26,179415,4.0,0.0,0.0
+2018-01-27,125769,5.0,0.0,0.0
+2018-01-28,133306,6.0,0.0,0.0
+2018-01-29,194151,0.0,0.0,0.0
+2018-01-30,198680,1.0,0.0,0.0
+2018-01-31,198652,2.0,0.0,0.0
+2018-02-01,195472,3.0,1.0,0.0
+2018-02-02,183173,4.0,1.0,0.0
+2018-02-03,124276,5.0,1.0,0.0
+2018-02-04,129054,6.0,1.0,0.0
+2018-02-05,190024,0.0,1.0,0.0
+2018-02-06,198658,1.0,1.0,0.0
+2018-02-07,198272,2.0,1.0,0.0
+2018-02-08,195339,3.0,1.0,0.0
+2018-02-09,183086,4.0,1.0,0.0
+2018-02-10,122536,5.0,1.0,0.0
+2018-02-11,133033,6.0,1.0,0.0
+2018-02-12,185386,0.0,1.0,0.0
+2018-02-13,184789,1.0,1.0,0.0
+2018-02-14,176089,2.0,1.0,0.0
+2018-02-15,171317,3.0,1.0,0.0
+2018-02-16,162693,4.0,1.0,0.0
+2018-02-17,116342,5.0,1.0,0.0
+2018-02-18,122466,6.0,1.0,0.0
+2018-02-19,172364,0.0,1.0,1.0
+2018-02-20,185896,1.0,1.0,0.0
+2018-02-21,188166,2.0,1.0,0.0
+2018-02-22,189427,3.0,1.0,0.0
+2018-02-23,178732,4.0,1.0,0.0
+2018-02-24,132664,5.0,1.0,0.0
+2018-02-25,134008,6.0,1.0,0.0
+2018-02-26,200075,0.0,1.0,0.0
+2018-02-27,207996,1.0,1.0,0.0
+2018-02-28,204416,2.0,1.0,0.0
+2018-03-01,201320,3.0,2.0,0.0
+2018-03-02,188205,4.0,2.0,0.0
+2018-03-03,131162,5.0,2.0,0.0
+2018-03-04,138320,6.0,2.0,0.0
+2018-03-05,207326,0.0,2.0,0.0
+2018-03-06,212462,1.0,2.0,0.0
+2018-03-07,209357,2.0,2.0,0.0
+2018-03-08,194876,3.0,2.0,0.0
+2018-03-09,193761,4.0,2.0,0.0
+2018-03-10,133449,5.0,2.0,0.0
+2018-03-11,142258,6.0,2.0,0.0
+2018-03-12,208753,0.0,2.0,0.0
+2018-03-13,210602,1.0,2.0,0.0
+2018-03-14,214236,2.0,2.0,0.0
+2018-03-15,210761,3.0,2.0,0.0
+2018-03-16,196619,4.0,2.0,0.0
+2018-03-17,133056,5.0,2.0,0.0
+2018-03-18,141335,6.0,2.0,0.0
+2018-03-19,211580,0.0,2.0,0.0
+2018-03-20,219051,1.0,2.0,0.0
+2018-03-21,215435,2.0,2.0,0.0
+2018-03-22,211961,3.0,2.0,0.0
+2018-03-23,196009,4.0,2.0,0.0
+2018-03-24,132390,5.0,2.0,0.0
+2018-03-25,140021,6.0,2.0,0.0
+2018-03-26,205273,0.0,2.0,0.0
+2018-03-27,212686,1.0,2.0,0.0
+2018-03-28,210683,2.0,2.0,0.0
+2018-03-29,189044,3.0,2.0,0.0
+2018-03-30,170256,4.0,2.0,0.0
+2018-03-31,125999,5.0,2.0,0.0
+2018-04-01,126749,6.0,3.0,0.0
+2018-04-02,186546,0.0,3.0,0.0
+2018-04-03,207905,1.0,3.0,0.0
+2018-04-04,201528,2.0,3.0,0.0
+2018-04-05,188580,3.0,3.0,0.0
+2018-04-06,173714,4.0,3.0,0.0
+2018-04-07,125723,5.0,3.0,0.0
+2018-04-08,142545,6.0,3.0,0.0
+2018-04-09,204767,0.0,3.0,0.0
+2018-04-10,212048,1.0,3.0,0.0
+2018-04-11,210517,2.0,3.0,0.0
+2018-04-12,206924,3.0,3.0,0.0
+2018-04-13,191679,4.0,3.0,0.0
+2018-04-14,126394,5.0,3.0,0.0
+2018-04-15,137279,6.0,3.0,0.0
+2018-04-16,208085,0.0,3.0,0.0
+2018-04-17,213273,1.0,3.0,0.0
+2018-04-18,211580,2.0,3.0,0.0
+2018-04-19,206037,3.0,3.0,0.0
+2018-04-20,191211,4.0,3.0,0.0
+2018-04-21,125564,5.0,3.0,0.0
+2018-04-22,136469,6.0,3.0,0.0
+2018-04-23,206288,0.0,3.0,0.0
+2018-04-24,212115,1.0,3.0,0.0
+2018-04-25,207948,2.0,3.0,0.0
+2018-04-26,205759,3.0,3.0,0.0
+2018-04-27,181330,4.0,3.0,0.0
+2018-04-28,130046,5.0,3.0,0.0
+2018-04-29,120802,6.0,3.0,0.0
+2018-04-30,170390,0.0,3.0,0.0
+2018-05-01,169054,1.0,4.0,0.0
+2018-05-02,197891,2.0,4.0,0.0
+2018-05-03,199820,3.0,4.0,0.0
+2018-05-04,186783,4.0,4.0,0.0
+2018-05-05,124420,5.0,4.0,0.0
+2018-05-06,130666,6.0,4.0,0.0
+2018-05-07,196014,0.0,4.0,0.0
+2018-05-08,203058,1.0,4.0,0.0
+2018-05-09,198582,2.0,4.0,0.0
+2018-05-10,191321,3.0,4.0,0.0
+2018-05-11,183639,4.0,4.0,0.0
+2018-05-12,122023,5.0,4.0,0.0
+2018-05-13,128775,6.0,4.0,0.0
+2018-05-14,199104,0.0,4.0,0.0
+2018-05-15,200658,1.0,4.0,0.0
+2018-05-16,201541,2.0,4.0,0.0
+2018-05-17,196886,3.0,4.0,0.0
+2018-05-18,188597,4.0,4.0,0.0
+2018-05-19,121392,5.0,4.0,0.0
+2018-05-20,126981,6.0,4.0,0.0
+2018-05-21,189291,0.0,4.0,0.0
+2018-05-22,203038,1.0,4.0,0.0
+2018-05-23,205330,2.0,4.0,0.0
+2018-05-24,199208,3.0,4.0,0.0
+2018-05-25,187768,4.0,4.0,0.0
+2018-05-26,117635,5.0,4.0,0.0
+2018-05-27,124352,6.0,4.0,0.0
+2018-05-28,180398,0.0,4.0,1.0
+2018-05-29,194170,1.0,4.0,0.0
+2018-05-30,200281,2.0,4.0,0.0
+2018-05-31,197244,3.0,4.0,0.0
+2018-06-01,184037,4.0,5.0,0.0
+2018-06-02,121135,5.0,5.0,0.0
+2018-06-03,129389,6.0,5.0,0.0
+2018-06-04,200331,0.0,5.0,0.0
+2018-06-05,207735,1.0,5.0,0.0
+2018-06-06,203354,2.0,5.0,0.0
+2018-06-07,200520,3.0,5.0,0.0
+2018-06-08,182038,4.0,5.0,0.0
+2018-06-09,120164,5.0,5.0,0.0
+2018-06-10,125256,6.0,5.0,0.0
+2018-06-11,194786,0.0,5.0,0.0
+2018-06-12,200815,1.0,5.0,0.0
+2018-06-13,197740,2.0,5.0,0.0
+2018-06-14,192294,3.0,5.0,0.0
+2018-06-15,173587,4.0,5.0,0.0
+2018-06-16,105955,5.0,5.0,0.0
+2018-06-17,110780,6.0,5.0,0.0
+2018-06-18,174582,0.0,5.0,0.0
+2018-06-19,193310,1.0,5.0,0.0
+2018-06-20,193062,2.0,5.0,0.0
+2018-06-21,187986,3.0,5.0,0.0
+2018-06-22,173606,4.0,5.0,0.0
+2018-06-23,111795,5.0,5.0,0.0
+2018-06-24,116134,6.0,5.0,0.0
+2018-06-25,185919,0.0,5.0,0.0
+2018-06-26,193142,1.0,5.0,0.0
+2018-06-27,188114,2.0,5.0,0.0
+2018-06-28,183737,3.0,5.0,0.0
+2018-06-29,171496,4.0,5.0,0.0
+2018-06-30,107210,5.0,5.0,0.0
+2018-07-01,111053,6.0,6.0,0.0
+2018-07-02,176198,0.0,6.0,0.0
+2018-07-03,184040,1.0,6.0,0.0
+2018-07-04,169783,2.0,6.0,1.0
+2018-07-05,177996,3.0,6.0,0.0
+2018-07-06,167378,4.0,6.0,0.0
+2018-07-07,106401,5.0,6.0,0.0
+2018-07-08,112327,6.0,6.0,0.0
+2018-07-09,182835,0.0,6.0,0.0
+2018-07-10,187694,1.0,6.0,0.0
+2018-07-11,185762,2.0,6.0,0.0
+2018-07-12,184099,3.0,6.0,0.0
+2018-07-13,170860,4.0,6.0,0.0
+2018-07-14,106799,5.0,6.0,0.0
+2018-07-15,108475,6.0,6.0,0.0
+2018-07-16,175704,0.0,6.0,0.0
+2018-07-17,183596,1.0,6.0,0.0
+2018-07-18,179897,2.0,6.0,0.0
+2018-07-19,183373,3.0,6.0,0.0
+2018-07-20,169626,4.0,6.0,0.0
+2018-07-21,106785,5.0,6.0,0.0
+2018-07-22,112387,6.0,6.0,0.0
+2018-07-23,180572,0.0,6.0,0.0
+2018-07-24,186943,1.0,6.0,0.0
+2018-07-25,185744,2.0,6.0,0.0
+2018-07-26,183117,3.0,6.0,0.0
+2018-07-27,168526,4.0,6.0,0.0
+2018-07-28,105936,5.0,6.0,0.0
+2018-07-29,111708,6.0,6.0,0.0
+2018-07-30,179950,0.0,6.0,0.0
+2018-07-31,185930,1.0,6.0,0.0
+2018-08-01,183366,2.0,7.0,0.0
+2018-08-02,182412,3.0,7.0,0.0
+2018-08-03,173429,4.0,7.0,0.0
+2018-08-04,106108,5.0,7.0,0.0
+2018-08-05,110059,6.0,7.0,0.0
+2018-08-06,178355,0.0,7.0,0.0
+2018-08-07,185518,1.0,7.0,0.0
+2018-08-08,183204,2.0,7.0,0.0
+2018-08-09,181276,3.0,7.0,0.0
+2018-08-10,168297,4.0,7.0,0.0
+2018-08-11,106488,5.0,7.0,0.0
+2018-08-12,111786,6.0,7.0,0.0
+2018-08-13,178620,0.0,7.0,0.0
+2018-08-14,181922,1.0,7.0,0.0
+2018-08-15,172198,2.0,7.0,0.0
+2018-08-16,177367,3.0,7.0,0.0
+2018-08-17,166550,4.0,7.0,0.0
+2018-08-18,107011,5.0,7.0,0.0
+2018-08-19,112299,6.0,7.0,0.0
+2018-08-20,176718,0.0,7.0,0.0
+2018-08-21,182562,1.0,7.0,0.0
+2018-08-22,181484,2.0,7.0,0.0
+2018-08-23,180317,3.0,7.0,0.0
+2018-08-24,170197,4.0,7.0,0.0
+2018-08-25,109383,5.0,7.0,0.0
+2018-08-26,113373,6.0,7.0,0.0
+2018-08-27,180142,0.0,7.0,0.0
+2018-08-28,191628,1.0,7.0,0.0
+2018-08-29,191149,2.0,7.0,0.0
+2018-08-30,187503,3.0,7.0,0.0
+2018-08-31,172280,4.0,7.0,0.0

how-to-use-azureml/automated-machine-learning/forecasting-github-dau/helper.py

@@ -0,0 +1,183 @@
+import pandas as pd
+from azureml.core import Environment
+from azureml.core.conda_dependencies import CondaDependencies
+from azureml.train.estimator import Estimator
+from azureml.core.run import Run
+from azureml.automl.core.shared import constants
+
+
+def split_fraction_by_grain(df, fraction, time_column_name, grain_column_names=None):
+    if not grain_column_names:
+        df["tmp_grain_column"] = "grain"
+        grain_column_names = ["tmp_grain_column"]
+
+    """Group df by grain and split on last n rows for each group."""
+    df_grouped = df.sort_values(time_column_name).groupby(
+        grain_column_names, group_keys=False
+    )
+
+    df_head = df_grouped.apply(
+        lambda dfg: dfg.iloc[: -int(len(dfg) * fraction)] if fraction > 0 else dfg
+    )
+
+    df_tail = df_grouped.apply(
+        lambda dfg: dfg.iloc[-int(len(dfg) * fraction) :] if fraction > 0 else dfg[:0]
+    )
+
+    if "tmp_grain_column" in grain_column_names:
+        for df2 in (df, df_head, df_tail):
+            df2.drop("tmp_grain_column", axis=1, inplace=True)
+
+        grain_column_names.remove("tmp_grain_column")
+
+    return df_head, df_tail
+
+
+def split_full_for_forecasting(
+    df, time_column_name, grain_column_names=None, test_split=0.2
+):
+    index_name = df.index.name
+
+    # Assumes that there isn't already a column called tmpindex
+
+    df["tmpindex"] = df.index
+
+    train_df, test_df = split_fraction_by_grain(
+        df, test_split, time_column_name, grain_column_names
+    )
+
+    train_df = train_df.set_index("tmpindex")
+    train_df.index.name = index_name
+
+    test_df = test_df.set_index("tmpindex")
+    test_df.index.name = index_name
+
+    df.drop("tmpindex", axis=1, inplace=True)
+
+    return train_df, test_df
+
+
+def get_result_df(remote_run):
+    children = list(remote_run.get_children(recursive=True))
+    summary_df = pd.DataFrame(
+        index=["run_id", "run_algorithm", "primary_metric", "Score"]
+    )
+    goal_minimize = False
+    for run in children:
+        if (
+            run.get_status().lower() == constants.RunState.COMPLETE_RUN
+            and "run_algorithm" in run.properties
+            and "score" in run.properties
+        ):
+            # We only count in the completed child runs.
+            summary_df[run.id] = [
+                run.id,
+                run.properties["run_algorithm"],
+                run.properties["primary_metric"],
+                float(run.properties["score"]),
+            ]
+            if "goal" in run.properties:
+                goal_minimize = run.properties["goal"].split("_")[-1] == "min"
+
+    summary_df = summary_df.T.sort_values(
+        "Score", ascending=goal_minimize
+    ).drop_duplicates(["run_algorithm"])
+    summary_df = summary_df.set_index("run_algorithm")
+    return summary_df
+
+
+def run_inference(
+    test_experiment,
+    compute_target,
+    script_folder,
+    train_run,
+    test_dataset,
+    lookback_dataset,
+    max_horizon,
+    target_column_name,
+    time_column_name,
+    freq,
+):
+    model_base_name = "model.pkl"
+    if "model_data_location" in train_run.properties:
+        model_location = train_run.properties["model_data_location"]
+        _, model_base_name = model_location.rsplit("/", 1)
+    train_run.download_file(
+        "outputs/{}".format(model_base_name), "inference/{}".format(model_base_name)
+    )
+    train_run.download_file("outputs/conda_env_v_1_0_0.yml", "inference/condafile.yml")
+
+    inference_env = Environment("myenv")
+    inference_env.docker.enabled = True
+    inference_env.python.conda_dependencies = CondaDependencies(
+        conda_dependencies_file_path="inference/condafile.yml"
+    )
+
+    est = Estimator(
+        source_directory=script_folder,
+        entry_script="infer.py",
+        script_params={
+            "--max_horizon": max_horizon,
+            "--target_column_name": target_column_name,
+            "--time_column_name": time_column_name,
+            "--frequency": freq,
+            "--model_path": model_base_name,
+        },
+        inputs=[
+            test_dataset.as_named_input("test_data"),
+            lookback_dataset.as_named_input("lookback_data"),
+        ],
+        compute_target=compute_target,
+        environment_definition=inference_env,
+    )
+
+    run = test_experiment.submit(
+        est,
+        tags={
+            "training_run_id": train_run.id,
+            "run_algorithm": train_run.properties["run_algorithm"],
+            "valid_score": train_run.properties["score"],
+            "primary_metric": train_run.properties["primary_metric"],
+        },
+    )
+
+    run.log("run_algorithm", run.tags["run_algorithm"])
+    return run
+
+
+def run_multiple_inferences(
+    summary_df,
+    train_experiment,
+    test_experiment,
+    compute_target,
+    script_folder,
+    test_dataset,
+    lookback_dataset,
+    max_horizon,
+    target_column_name,
+    time_column_name,
+    freq,
+):
+    for run_name, run_summary in summary_df.iterrows():
+        print(run_name)
+        print(run_summary)
+        run_id = run_summary.run_id
+        train_run = Run(train_experiment, run_id)
+
+        test_run = run_inference(
+            test_experiment,
+            compute_target,
+            script_folder,
+            train_run,
+            test_dataset,
+            lookback_dataset,
+            max_horizon,
+            target_column_name,
+            time_column_name,
+            freq,
+        )
+
+        print(test_run)
+        summary_df.loc[summary_df.run_id == run_id, "test_run_id"] = test_run.id
+
+    return summary_df

how-to-use-azureml/automated-machine-learning/forecasting-github-dau/infer.py

@@ -0,0 +1,386 @@
+import argparse
+import os
+
+import numpy as np
+import pandas as pd
+
+from pandas.tseries.frequencies import to_offset
+from sklearn.externals import joblib
+from sklearn.metrics import mean_absolute_error, mean_squared_error
+
+from azureml.automl.runtime.shared.score import scoring, constants
+from azureml.core import Run
+
+try:
+    import torch
+
+    _torch_present = True
+except ImportError:
+    _torch_present = False
+
+
+def align_outputs(
+    y_predicted,
+    X_trans,
+    X_test,
+    y_test,
+    predicted_column_name="predicted",
+    horizon_colname="horizon_origin",
+):
+    """
+    Demonstrates how to get the output aligned to the inputs
+    using pandas indexes. Helps understand what happened if
+    the output's shape differs from the input shape, or if
+    the data got re-sorted by time and grain during forecasting.
+
+    Typical causes of misalignment are:
+    * we predicted some periods that were missing in actuals -> drop from eval
+    * model was asked to predict past max_horizon -> increase max horizon
+    * data at start of X_test was needed for lags -> provide previous periods
+    """
+    if horizon_colname in X_trans:
+        df_fcst = pd.DataFrame(
+            {
+                predicted_column_name: y_predicted,
+                horizon_colname: X_trans[horizon_colname],
+            }
+        )
+    else:
+        df_fcst = pd.DataFrame({predicted_column_name: y_predicted})
+
+    # y and X outputs are aligned by forecast() function contract
+    df_fcst.index = X_trans.index
+
+    # align original X_test to y_test
+    X_test_full = X_test.copy()
+    X_test_full[target_column_name] = y_test
+
+    # X_test_full's index does not include origin, so reset for merge
+    df_fcst.reset_index(inplace=True)
+    X_test_full = X_test_full.reset_index().drop(columns="index")
+    together = df_fcst.merge(X_test_full, how="right")
+
+    # drop rows where prediction or actuals are nan
+    # happens because of missing actuals
+    # or at edges of time due to lags/rolling windows
+    clean = together[
+        together[[target_column_name, predicted_column_name]].notnull().all(axis=1)
+    ]
+    return clean
+
+
+def do_rolling_forecast_with_lookback(
+    fitted_model, X_test, y_test, max_horizon, X_lookback, y_lookback, freq="D"
+):
+    """
+    Produce forecasts on a rolling origin over the given test set.
+
+    Each iteration makes a forecast for the next 'max_horizon' periods
+    with respect to the current origin, then advances the origin by the
+    horizon time duration. The prediction context for each forecast is set so
+    that the forecaster uses the actual target values prior to the current
+    origin time for constructing lag features.
+
+    This function returns a concatenated DataFrame of rolling forecasts.
+    """
+    print("Using lookback of size: ", y_lookback.size)
+    df_list = []
+    origin_time = X_test[time_column_name].min()
+    X = X_lookback.append(X_test)
+    y = np.concatenate((y_lookback, y_test), axis=0)
+    while origin_time <= X_test[time_column_name].max():
+        # Set the horizon time - end date of the forecast
+        horizon_time = origin_time + max_horizon * to_offset(freq)
+
+        # Extract test data from an expanding window up-to the horizon
+        expand_wind = X[time_column_name] < horizon_time
+        X_test_expand = X[expand_wind]
+        y_query_expand = np.zeros(len(X_test_expand)).astype(np.float)
+        y_query_expand.fill(np.NaN)
+
+        if origin_time != X[time_column_name].min():
+            # Set the context by including actuals up-to the origin time
+            test_context_expand_wind = X[time_column_name] < origin_time
+            context_expand_wind = X_test_expand[time_column_name] < origin_time
+            y_query_expand[context_expand_wind] = y[test_context_expand_wind]
+
+        # Print some debug info
+        print(
+            "Horizon_time:",
+            horizon_time,
+            " origin_time: ",
+            origin_time,
+            " max_horizon: ",
+            max_horizon,
+            " freq: ",
+            freq,
+        )
+        print("expand_wind: ", expand_wind)
+        print("y_query_expand")
+        print(y_query_expand)
+        print("X_test")
+        print(X)
+        print("X_test_expand")
+        print(X_test_expand)
+        print("Type of X_test_expand: ", type(X_test_expand))
+        print("Type of y_query_expand: ", type(y_query_expand))
+
+        print("y_query_expand")
+        print(y_query_expand)
+
+        # Make a forecast out to the maximum horizon
+        # y_fcst, X_trans = y_query_expand, X_test_expand
+        y_fcst, X_trans = fitted_model.forecast(X_test_expand, y_query_expand)
+
+        print("y_fcst")
+        print(y_fcst)
+
+        # Align forecast with test set for dates within
+        # the current rolling window
+        trans_tindex = X_trans.index.get_level_values(time_column_name)
+        trans_roll_wind = (trans_tindex >= origin_time) & (trans_tindex < horizon_time)
+        test_roll_wind = expand_wind & (X[time_column_name] >= origin_time)
+        df_list.append(
+            align_outputs(
+                y_fcst[trans_roll_wind],
+                X_trans[trans_roll_wind],
+                X[test_roll_wind],
+                y[test_roll_wind],
+            )
+        )
+
+        # Advance the origin time
+        origin_time = horizon_time
+
+    return pd.concat(df_list, ignore_index=True)
+
+
+def do_rolling_forecast(fitted_model, X_test, y_test, max_horizon, freq="D"):
+    """
+    Produce forecasts on a rolling origin over the given test set.
+
+    Each iteration makes a forecast for the next 'max_horizon' periods
+    with respect to the current origin, then advances the origin by the
+    horizon time duration. The prediction context for each forecast is set so
+    that the forecaster uses the actual target values prior to the current
+    origin time for constructing lag features.
+
+    This function returns a concatenated DataFrame of rolling forecasts.
+    """
+    df_list = []
+    origin_time = X_test[time_column_name].min()
+    while origin_time <= X_test[time_column_name].max():
+        # Set the horizon time - end date of the forecast
+        horizon_time = origin_time + max_horizon * to_offset(freq)
+
+        # Extract test data from an expanding window up-to the horizon
+        expand_wind = X_test[time_column_name] < horizon_time
+        X_test_expand = X_test[expand_wind]
+        y_query_expand = np.zeros(len(X_test_expand)).astype(np.float)
+        y_query_expand.fill(np.NaN)
+
+        if origin_time != X_test[time_column_name].min():
+            # Set the context by including actuals up-to the origin time
+            test_context_expand_wind = X_test[time_column_name] < origin_time
+            context_expand_wind = X_test_expand[time_column_name] < origin_time
+            y_query_expand[context_expand_wind] = y_test[test_context_expand_wind]
+
+        # Print some debug info
+        print(
+            "Horizon_time:",
+            horizon_time,
+            " origin_time: ",
+            origin_time,
+            " max_horizon: ",
+            max_horizon,
+            " freq: ",
+            freq,
+        )
+        print("expand_wind: ", expand_wind)
+        print("y_query_expand")
+        print(y_query_expand)
+        print("X_test")
+        print(X_test)
+        print("X_test_expand")
+        print(X_test_expand)
+        print("Type of X_test_expand: ", type(X_test_expand))
+        print("Type of y_query_expand: ", type(y_query_expand))
+        print("y_query_expand")
+        print(y_query_expand)
+
+        # Make a forecast out to the maximum horizon
+        y_fcst, X_trans = fitted_model.forecast(X_test_expand, y_query_expand)
+
+        print("y_fcst")
+        print(y_fcst)
+
+        # Align forecast with test set for dates within the
+        # current rolling window
+        trans_tindex = X_trans.index.get_level_values(time_column_name)
+        trans_roll_wind = (trans_tindex >= origin_time) & (trans_tindex < horizon_time)
+        test_roll_wind = expand_wind & (X_test[time_column_name] >= origin_time)
+        df_list.append(
+            align_outputs(
+                y_fcst[trans_roll_wind],
+                X_trans[trans_roll_wind],
+                X_test[test_roll_wind],
+                y_test[test_roll_wind],
+            )
+        )
+
+        # Advance the origin time
+        origin_time = horizon_time
+
+    return pd.concat(df_list, ignore_index=True)
+
+
+def APE(actual, pred):
+    """
+    Calculate absolute percentage error.
+    Returns a vector of APE values with same length as actual/pred.
+    """
+    return 100 * np.abs((actual - pred) / actual)
+
+
+def MAPE(actual, pred):
+    """
+    Calculate mean absolute percentage error.
+    Remove NA and values where actual is close to zero
+    """
+    not_na = ~(np.isnan(actual) | np.isnan(pred))
+    not_zero = ~np.isclose(actual, 0.0)
+    actual_safe = actual[not_na & not_zero]
+    pred_safe = pred[not_na & not_zero]
+    return np.mean(APE(actual_safe, pred_safe))
+
+
+def map_location_cuda(storage, loc):
+    return storage.cuda()
+
+
+parser = argparse.ArgumentParser()
+parser.add_argument(
+    "--max_horizon",
+    type=int,
+    dest="max_horizon",
+    default=10,
+    help="Max Horizon for forecasting",
+)
+parser.add_argument(
+    "--target_column_name",
+    type=str,
+    dest="target_column_name",
+    help="Target Column Name",
+)
+parser.add_argument(
+    "--time_column_name", type=str, dest="time_column_name", help="Time Column Name"
+)
+parser.add_argument(
+    "--frequency", type=str, dest="freq", help="Frequency of prediction"
+)
+parser.add_argument(
+    "--model_path",
+    type=str,
+    dest="model_path",
+    default="model.pkl",
+    help="Filename of model to be loaded",
+)
+
+args = parser.parse_args()
+max_horizon = args.max_horizon
+target_column_name = args.target_column_name
+time_column_name = args.time_column_name
+freq = args.freq
+model_path = args.model_path
+
+print("args passed are: ")
+print(max_horizon)
+print(target_column_name)
+print(time_column_name)
+print(freq)
+print(model_path)
+
+run = Run.get_context()
+# get input dataset by name
+test_dataset = run.input_datasets["test_data"]
+lookback_dataset = run.input_datasets["lookback_data"]
+
+grain_column_names = []
+
+df = test_dataset.to_pandas_dataframe()
+
+print("Read df")
+print(df)
+
+X_test_df = test_dataset.drop_columns(columns=[target_column_name])
+y_test_df = test_dataset.with_timestamp_columns(None).keep_columns(
+    columns=[target_column_name]
+)
+
+X_lookback_df = lookback_dataset.drop_columns(columns=[target_column_name])
+y_lookback_df = lookback_dataset.with_timestamp_columns(None).keep_columns(
+    columns=[target_column_name]
+)
+
+_, ext = os.path.splitext(model_path)
+if ext == ".pt":
+    # Load the fc-tcn torch model.
+    assert _torch_present
+    if torch.cuda.is_available():
+        map_location = map_location_cuda
+    else:
+        map_location = "cpu"
+    with open(model_path, "rb") as fh:
+        fitted_model = torch.load(fh, map_location=map_location)
+else:
+    # Load the sklearn pipeline.
+    fitted_model = joblib.load(model_path)
+
+if hasattr(fitted_model, "get_lookback"):
+    lookback = fitted_model.get_lookback()
+    df_all = do_rolling_forecast_with_lookback(
+        fitted_model,
+        X_test_df.to_pandas_dataframe(),
+        y_test_df.to_pandas_dataframe().values.T[0],
+        max_horizon,
+        X_lookback_df.to_pandas_dataframe()[-lookback:],
+        y_lookback_df.to_pandas_dataframe().values.T[0][-lookback:],
+        freq,
+    )
+else:
+    df_all = do_rolling_forecast(
+        fitted_model,
+        X_test_df.to_pandas_dataframe(),
+        y_test_df.to_pandas_dataframe().values.T[0],
+        max_horizon,
+        freq,
+    )
+
+print(df_all)
+
+print("target values:::")
+print(df_all[target_column_name])
+print("predicted values:::")
+print(df_all["predicted"])
+
+# Use the AutoML scoring module
+regression_metrics = list(constants.REGRESSION_SCALAR_SET)
+y_test = np.array(df_all[target_column_name])
+y_pred = np.array(df_all["predicted"])
+scores = scoring.score_regression(y_test, y_pred, regression_metrics)
+
+print("scores:")
+print(scores)
+
+for key, value in scores.items():
+    run.log(key, value)
+
+print("Simple forecasting model")
+rmse = np.sqrt(mean_squared_error(df_all[target_column_name], df_all["predicted"]))
+print("[Test Data] \nRoot Mean squared error: %.2f" % rmse)
+mae = mean_absolute_error(df_all[target_column_name], df_all["predicted"])
+print("mean_absolute_error score: %.2f" % mae)
+print("MAPE: %.2f" % MAPE(df_all[target_column_name], df_all["predicted"]))
+
+run.log("rmse", rmse)
+run.log("mae", mae)

how-to-use-azureml/automated-machine-learning/forecasting-hierarchical-timeseries/README.md

@@ -0,0 +1,94 @@
+---
+page_type: sample
+languages:
+- python
+products:
+- azure-machine-learning
+description: Tutorial showing how to solve a complex machine learning time series forecasting problems at scale by using Azure Automated ML and Hierarchical time series accelerator.
+---
+
+## Microsoft Solution Accelerator: Hierachical Time Series Forecasting
+
+In most applications, customers have a need to understand their forecasts at a macro and micro level of the business. Whether that be predicting sales of products at different geographic locations, or understanding the expected workforce demand for different organizations at a company, the ability to train a machine learning model to intelligently forecast on hierarchy data is essential.
+
+This business pattern is common across a wide variety of industries and applicable to many real world use cases. Below are some examples of where the hierarchical time series pattern is useful.
+
+| Industry       | Scenario                                |
+|----------------|--------------------------------------------|
+| *Restaurant Chain* | Building demand forecasting models across thousands of restaurants and several countries. |
+| *Retail Organization* | Building workforce optimization models for thousands of stores. |
+| *Retail Organization*| Price optimization models for hundreds of thousands of products available. |
+
+
+### Technical Summary
+
+A hierarchical time series is a structure in which each of the unique series are arranged into a hierarchy based on dimensions such as geography, or product type. The table below shows an example of data whose unique attributes form a hierarchy. Our hierarchy is defined by the `product type` such as headphones or tablets, the `product category` which splits product types into accessories and devices, and the `region` the products are sold in. The table below demonstrates the first input of each unique series in the hierarchy.
+
+![data-table](./media/data-table.png)
+
+To further visualize this, the leaf levels of the hierarchy contain all the time series with unique combinations of attribute values. Each higher level in the hierarchy will consider one less dimension for defining the time series and will aggregate each set of `child nodes` from the lower level into a `parent node`.
+
+![hierachy-sample](./media/hierarchy-sample-ms.PNG)
+
+> **Note:** If no unique root level exists in the data, Automated Machine Learning will create a node `automl_top_level` for users to train or forecasts totals.
+
+## Prerequisites
+
+To use this solution accelerator, all you need is access to an [Azure subscription](https://azure.microsoft.com/free/) and an [Azure Machine Learning Workspace](https://docs.microsoft.com/azure/machine-learning/how-to-manage-workspace) that you'll create below.
+
+A basic understanding of Azure Machine Learning and hierarchical time series concepts will be helpful for understanding the solution. The following resources can help introduce you to these concepts:
+
+1. [Azure Machine Learning Overview](https://azure.microsoft.com/services/machine-learning/)
+2. [Azure Machine Learning Tutorials](https://docs.microsoft.com/azure/machine-learning/tutorial-1st-experiment-sdk-setup)
+3. [Azure Machine Learning Sample Notebooks on Github](https://github.com/Azure/azureml-examples/)
+4. [Forecasting: Principles and Practice, Hierarchical time series](https://otexts.com/fpp2/hts.html)
+
+## Getting started
+
+### 1. Set up the Compute Instance
+Please create a [Compute Instance](https://docs.microsoft.com/en-us/azure/machine-learning/concept-compute-instance#create) and clone the git repo to your workspace.
+
+### 2. Run the Notebook
+
+Once your environment is set up, go to JupyterLab and run the notebook auto-ml-hierarchical-timeseries.ipynb on Compute Instance you created. It would run through the steps outlined sequentially. By the end, you'll know how to train, score, and make predictions using the hierarchical time series model pattern on Azure Machine Learning.
+
+| Notebook       | Description                                |
+|----------------|--------------------------------------------|
+| `auto-ml-forecasting-hierarchical-timeseries.ipynb`|Creates a pipeline to train machine learning models for the defined hierarchy and forecast at the desired hierarchy level using Automated ML. |
+
+
+![Work Flow](./media/workflow.PNG)
+
+## Key Concepts
+
+### Automated Machine Learning
+
+[Automated Machine Learning](https://docs.microsoft.com/azure/machine-learning/concept-automated-ml) also referred to as automated ML or AutoML, is the process of automating the time consuming, iterative tasks of machine learning model development. It allows data scientists, analysts, and developers to build ML models with high scale, efficiency, and productivity all while sustaining model quality.
+
+### Pipelines
+
+[Pipelines](https://docs.microsoft.com/azure/machine-learning/concept-ml-pipelines) allow you to create workflows in your machine learning projects. These workflows have a number of benefits including speed, simplicity, repeatability, and modularity.
+
+### ParallelRunStep
+
+[ParallelRunStep](https://docs.microsoft.com/en-us/python/api/azureml-pipeline-steps/azureml.pipeline.steps.parallel_run_step.parallelrunstep?view=azure-ml-py) enables the parallel training of models and is commonly used for batch inferencing. This [document](https://docs.microsoft.com/azure/machine-learning/how-to-use-parallel-run-step) walks through some of the key concepts around ParallelRunStep.
+
+### Other Concepts
+
+In additional to ParallelRunStep, Pipelines and Automated Machine Learning, you'll also be working with the following concepts including [workspace](https://docs.microsoft.com/azure/machine-learning/concept-workspace), [datasets](https://docs.microsoft.com/azure/machine-learning/concept-data#datasets), [compute targets](https://docs.microsoft.com/azure/machine-learning/concept-compute-target#train), [python script steps](https://docs.microsoft.com/python/api/azureml-pipeline-steps/azureml.pipeline.steps.python_script_step.pythonscriptstep?view=azure-ml-py), and [Azure Open Datasets](https://azure.microsoft.com/services/open-datasets/).
+
+## Contributing
+
+This project welcomes contributions and suggestions. To learn more visit the [contributing](CONTRIBUTING.md) section.
+
+Most contributions require you to agree to a Contributor License Agreement (CLA)
+declaring that you have the right to, and actually do, grant us
+the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.
+
+When you submit a pull request, a CLA bot will automatically determine whether you need to provide
+a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions
+provided by the bot. You will only need to do this once across all repos using our CLA.
+
+This project has adopted the [Microsoft Open Source Code of Conduct](https://opensource.microsoft.com/codeofconduct/).
+For more information see the [Code of Conduct FAQ](https://opensource.microsoft.com/codeofconduct/faq/) or
+contact [opencode@microsoft.com](mailto:opencode@microsoft.com) with any additional questions or comments.

how-to-use-azureml/automated-machine-learning/forecasting-hierarchical-timeseries/auto-ml-forecasting-hierarchical-timeseries.ipynb

@@ -0,0 +1,639 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Copyright (c) Microsoft Corporation. All rights reserved.\n",
+    "\n",
+    "Licensed under the MIT License."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "![Impressions](https://PixelServer20190423114238.azurewebsites.net/api/impressions/MachineLearningNotebooks/how-to-use-azureml/automated-machine-learning/forecasting-hierarchical-timeseries/auto-ml-forecasting-hierarchical-timeseries.png)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Hierarchical Time Series - Automated ML\n",
+    "**_Generate hierarchical time series forecasts with Automated Machine Learning_**\n",
+    "\n",
+    "---"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "For this notebook we are using a synthetic dataset portraying sales data to predict the the quantity of a vartiety of product skus across several states, stores, and product categories.\n",
+    "\n",
+    "**NOTE: There are limits on how many runs we can do in parallel per workspace, and we currently recommend to set the parallelism to maximum of 320 runs per experiment per workspace. If users want to have more parallelism and increase this limit they might encounter Too Many Requests errors (HTTP 429).**"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Prerequisites\n",
+    "You'll need to create a compute Instance by following the instructions in the [EnvironmentSetup.md](../Setup_Resources/EnvironmentSetup.md)."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 1.0 Set up workspace, datastore, experiment"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613003526897
+    }
+   },
+   "outputs": [],
+   "source": [
+    "import azureml.core\n",
+    "from azureml.core import Workspace, Datastore\n",
+    "import pandas as pd\n",
+    "\n",
+    "# Set up your workspace\n",
+    "ws = Workspace.from_config()\n",
+    "ws.get_details()\n",
+    "\n",
+    "# Set up your datastores\n",
+    "dstore = ws.get_default_datastore()\n",
+    "\n",
+    "output = {}\n",
+    "output[\"SDK version\"] = azureml.core.VERSION\n",
+    "output[\"Subscription ID\"] = ws.subscription_id\n",
+    "output[\"Workspace\"] = ws.name\n",
+    "output[\"Resource Group\"] = ws.resource_group\n",
+    "output[\"Location\"] = ws.location\n",
+    "output[\"Default datastore name\"] = dstore.name\n",
+    "pd.set_option(\"display.max_colwidth\", -1)\n",
+    "outputDf = pd.DataFrame(data=output, index=[\"\"])\n",
+    "outputDf.T"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Choose an experiment"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613003540729
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.core import Experiment\n",
+    "\n",
+    "experiment = Experiment(ws, \"automl-hts\")\n",
+    "\n",
+    "print(\"Experiment name: \" + experiment.name)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 2.0 Data\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "nteract": {
+     "transient": {
+      "deleting": false
+     }
+    }
+   },
+   "source": [
+    "### Upload local csv files to datastore\n",
+    "You can upload your train and inference csv files to the default datastore in your workspace. \n",
+    "\n",
+    "A Datastore is a place where data can be stored that is then made accessible to a compute either by means of mounting or copying the data to the compute target.\n",
+    "Please refer to [Datastore](https://docs.microsoft.com/en-us/python/api/azureml-core/azureml.core.datastore.datastore?view=azure-ml-py) documentation on how to access data from Datastore."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "datastore_path = \"hts-sample\""
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "datastore = ws.get_default_datastore()\n",
+    "datastore"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Create the TabularDatasets \n",
+    "\n",
+    "Datasets in Azure Machine Learning are references to specific data in a Datastore. The data can be retrieved as a [TabularDatasets](https://docs.microsoft.com/en-us/python/api/azureml-core/azureml.data.tabulardataset?view=azure-ml-py). We will read in the data as a pandas DataFrame, upload to the data store and register them to your Workspace using ```register_pandas_dataframe``` so they can be called as an input into the training pipeline. We will use the inference dataset as part of the forecasting pipeline. The step need only be completed once."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613007017296
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.data.dataset_factory import TabularDatasetFactory\n",
+    "\n",
+    "registered_train = TabularDatasetFactory.register_pandas_dataframe(\n",
+    "    pd.read_csv(\"Data/hts-sample-train.csv\"),\n",
+    "    target=(datastore, \"hts-sample\"),\n",
+    "    name=\"hts-sales-train\",\n",
+    ")\n",
+    "registered_inference = TabularDatasetFactory.register_pandas_dataframe(\n",
+    "    pd.read_csv(\"Data/hts-sample-test.csv\"),\n",
+    "    target=(datastore, \"hts-sample\"),\n",
+    "    name=\"hts-sales-test\",\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 3.0 Build the training pipeline\n",
+    "Now that the dataset, WorkSpace, and datastore are set up, we can put together a pipeline for training.\n",
+    "\n",
+    "> Note that if you have an AzureML Data Scientist role, you will not have permission to create compute resources. Talk to your workspace or IT admin to create the compute targets described in this section, if they do not already exist."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Choose a compute target\n",
+    "\n",
+    "You will need to create a [compute target](https://docs.microsoft.com/en-us/azure/machine-learning/how-to-set-up-training-targets#amlcompute) for your AutoML run. In this tutorial, you create AmlCompute as your training compute resource.\n",
+    "\n",
+    "\\*\\*Creation of AmlCompute takes approximately 5 minutes.**\n",
+    "\n",
+    "If the AmlCompute with that name is already in your workspace this code will skip the creation process. As with other Azure services, there are limits on certain resources (e.g. AmlCompute) associated with the Azure Machine Learning service. Please read this [article](https://docs.microsoft.com/en-us/azure/machine-learning/how-to-manage-quotas) on the default limits and how to request more quota."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613007037308
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.core.compute import ComputeTarget, AmlCompute\n",
+    "\n",
+    "# Name your cluster\n",
+    "compute_name = \"hts-compute\"\n",
+    "\n",
+    "\n",
+    "if compute_name in ws.compute_targets:\n",
+    "    compute_target = ws.compute_targets[compute_name]\n",
+    "    if compute_target and type(compute_target) is AmlCompute:\n",
+    "        print(\"Found compute target: \" + compute_name)\n",
+    "else:\n",
+    "    print(\"Creating a new compute target...\")\n",
+    "    provisioning_config = AmlCompute.provisioning_configuration(\n",
+    "        vm_size=\"STANDARD_D16S_V3\", max_nodes=20\n",
+    "    )\n",
+    "    # Create the compute target\n",
+    "    compute_target = ComputeTarget.create(ws, compute_name, provisioning_config)\n",
+    "\n",
+    "    # Can poll for a minimum number of nodes and for a specific timeout.\n",
+    "    # If no min node count is provided it will use the scale settings for the cluster\n",
+    "    compute_target.wait_for_completion(\n",
+    "        show_output=True, min_node_count=None, timeout_in_minutes=20\n",
+    "    )\n",
+    "\n",
+    "    # For a more detailed view of current cluster status, use the 'status' property\n",
+    "    print(compute_target.status.serialize())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Set up training parameters\n",
+    "\n",
+    "This dictionary defines the AutoML and hierarchy settings. For this forecasting task we need to define several settings inncluding the name of the time column, the maximum forecast horizon, the hierarchy definition, and the level of the hierarchy at which to train.\n",
+    "\n",
+    "| Property                           | Description|\n",
+    "| :---------------                   | :------------------- |\n",
+    "| **task**                           | forecasting |\n",
+    "| **primary_metric**                 | This is the metric that you want to optimize.<br> Forecasting supports the following primary metrics <br><i>spearman_correlation</i><br><i>normalized_root_mean_squared_error</i><br><i>r2_score</i><br><i>normalized_mean_absolute_error</i> |\n",
+    "| **blocked_models**                 | Blocked models won't be used by AutoML. |\n",
+    "| **iteration_timeout_minutes**      | Maximum amount of time in minutes that the model can train. This is optional but provides customers with greater control on exit criteria. |\n",
+    "| **iterations**                     | Number of models to train. This is optional but provides customers with greater control on exit criteria. |\n",
+    "| **experiment_timeout_hours**       | Maximum amount of time in hours that the experiment can take before it terminates. This is optional but provides customers with greater control on exit criteria. |\n",
+    "| **label_column_name**              | The name of the label column. |\n",
+    "| **forecast_horizon**               | The forecast horizon is how many periods forward you would like to forecast. This integer horizon is in units of the timeseries frequency (e.g. daily, weekly). Periods are inferred from your data. |\n",
+    "| **n_cross_validations**            | Number of cross validation splits. Rolling Origin Validation is used to split time-series in a temporally consistent way. |\n",
+    "| **enable_early_stopping**          | Flag to enable early termination if the score is not improving in the short term. |\n",
+    "| **time_column_name**               | The name of your time column. |\n",
+    "| **hierarchy_column_names**         | The names of columns that define the hierarchical structure of the data from highest level to most granular. |\n",
+    "| **training_level**                 | The level of the hierarchy to be used for training models. |\n",
+    "| **enable_engineered_explanations** | Engineered feature explanations will be downloaded if enable_engineered_explanations flag is set to True. By default it is set to False to save storage space. |\n",
+    "| **time_series_id_column_name**     | The column names used to uniquely identify timeseries in data that has multiple rows with the same timestamp. |\n",
+    "| **track_child_runs**               | Flag to disable tracking of child runs. Only best run is tracked if the flag is set to False (this includes the model and metrics of the run). |\n",
+    "| **pipeline_fetch_max_batch_size**  | Determines how many pipelines (training algorithms) to fetch at a time for training, this helps reduce throttling when training at large scale. |\n",
+    "| **model_explainability**           | Flag to disable explaining the best automated ML model at the end of all training iterations. The default is True and will block non-explainable models which may impact the forecast accuracy. For more information, see [Interpretability: model explanations in automated machine learning](https://docs.microsoft.com/en-us/azure/machine-learning/how-to-machine-learning-interpretability-automl). |"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613007061544
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.train.automl.runtime._hts.hts_parameters import HTSTrainParameters\n",
+    "\n",
+    "model_explainability = True\n",
+    "\n",
+    "engineered_explanations = False\n",
+    "# Define your hierarchy. Adjust the settings below based on your dataset.\n",
+    "hierarchy = [\"state\", \"store_id\", \"product_category\", \"SKU\"]\n",
+    "training_level = \"SKU\"\n",
+    "\n",
+    "# Set your forecast parameters. Adjust the settings below based on your dataset.\n",
+    "time_column_name = \"date\"\n",
+    "label_column_name = \"quantity\"\n",
+    "forecast_horizon = 7\n",
+    "\n",
+    "\n",
+    "automl_settings = {\n",
+    "    \"task\": \"forecasting\",\n",
+    "    \"primary_metric\": \"normalized_root_mean_squared_error\",\n",
+    "    \"label_column_name\": label_column_name,\n",
+    "    \"time_column_name\": time_column_name,\n",
+    "    \"forecast_horizon\": forecast_horizon,\n",
+    "    \"hierarchy_column_names\": hierarchy,\n",
+    "    \"hierarchy_training_level\": training_level,\n",
+    "    \"track_child_runs\": False,\n",
+    "    \"pipeline_fetch_max_batch_size\": 15,\n",
+    "    \"model_explainability\": model_explainability,\n",
+    "    # The following settings are specific to this sample and should be adjusted according to your own needs.\n",
+    "    \"iteration_timeout_minutes\": 10,\n",
+    "    \"iterations\": 10,\n",
+    "    \"n_cross_validations\": 2,\n",
+    "}\n",
+    "\n",
+    "hts_parameters = HTSTrainParameters(\n",
+    "    automl_settings=automl_settings,\n",
+    "    hierarchy_column_names=hierarchy,\n",
+    "    training_level=training_level,\n",
+    "    enable_engineered_explanations=engineered_explanations,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Set up hierarchy training pipeline"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Parallel run step is leveraged to train the hierarchy. To configure the ParallelRunConfig you will need to determine the appropriate number of workers and nodes for your use case. The `process_count_per_node` is based off the number of cores of the compute VM. The node_count will determine the number of master nodes to use, increasing the node count will speed up the training process.\n",
+    "\n",
+    "* **experiment:** The experiment used for training.\n",
+    "* **train_data:** The tabular dataset to be used as input to the training run.\n",
+    "* **node_count:** The number of compute nodes to be used for running the user script. We recommend to start with 3 and increase the node_count if the training time is taking too long.\n",
+    "* **process_count_per_node:** Process count per node, we recommend 2:1 ratio for number of cores: number of processes per node. eg. If node has 16 cores then configure 8 or less process count per node or optimal performance.\n",
+    "* **train_pipeline_parameters:** The set of configuration parameters defined in the previous section. \n",
+    "\n",
+    "Calling this method will create a new aggregated dataset which is generated dynamically on pipeline execution."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.contrib.automl.pipeline.steps import AutoMLPipelineBuilder\n",
+    "\n",
+    "\n",
+    "training_pipeline_steps = AutoMLPipelineBuilder.get_many_models_train_steps(\n",
+    "    experiment=experiment,\n",
+    "    train_data=registered_train,\n",
+    "    compute_target=compute_target,\n",
+    "    node_count=2,\n",
+    "    process_count_per_node=8,\n",
+    "    train_pipeline_parameters=hts_parameters,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.pipeline.core import Pipeline\n",
+    "\n",
+    "training_pipeline = Pipeline(ws, steps=training_pipeline_steps)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Submit the pipeline to run\n",
+    "Next we submit our pipeline to run. The whole training pipeline takes about 1h using a Standard_D16_V3 VM with our current ParallelRunConfig setting."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "training_run = experiment.submit(training_pipeline)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "training_run.wait_for_completion(show_output=True)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Check the run status, if training_run is in completed state, continue to forecasting. If training_run is in another state, check the portal for failures."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### [Optional] Get the explanations\n",
+    "First we need to download the explanations to the local disk."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "if model_explainability:\n",
+    "    expl_output = training_run.get_pipeline_output(\"explanations\")\n",
+    "    expl_output.download(\"training_explanations\")\n",
+    "else:\n",
+    "    print(\n",
+    "        \"Model explanations are available only if model_explainability is set to True.\"\n",
+    "    )"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "The explanations are downloaded to the \"training_explanations/azureml\" directory."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "\n",
+    "if model_explainability:\n",
+    "    explanations_dirrectory = os.listdir(\n",
+    "        os.path.join(\"training_explanations\", \"azureml\")\n",
+    "    )\n",
+    "    if len(explanations_dirrectory) > 1:\n",
+    "        print(\n",
+    "            \"Warning! The directory contains multiple explanations, only the first one will be displayed.\"\n",
+    "        )\n",
+    "    print(\"The explanations are located at {}.\".format(explanations_dirrectory[0]))\n",
+    "    # Now we will list all the explanations.\n",
+    "    explanation_path = os.path.join(\n",
+    "        \"training_explanations\",\n",
+    "        \"azureml\",\n",
+    "        explanations_dirrectory[0],\n",
+    "        \"training_explanations\",\n",
+    "    )\n",
+    "    print(\"Available explanations\")\n",
+    "    print(\"==============================\")\n",
+    "    print(\"\\n\".join(os.listdir(explanation_path)))\n",
+    "else:\n",
+    "    print(\n",
+    "        \"Model explanations are available only if model_explainability is set to True.\"\n",
+    "    )"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "View the explanations on \"state\" level."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from IPython.display import display\n",
+    "\n",
+    "explanation_type = \"raw\"\n",
+    "level = \"state\"\n",
+    "\n",
+    "if model_explainability:\n",
+    "    display(\n",
+    "        pd.read_csv(\n",
+    "            os.path.join(explanation_path, \"{}_explanations_{}.csv\").format(\n",
+    "                explanation_type, level\n",
+    "            )\n",
+    "        )\n",
+    "    )"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 5.0 Forecasting\n",
+    "For hierarchical forecasting we need to provide the HTSInferenceParameters object.\n",
+    "#### HTSInferenceParameters arguments\n",
+    "* **hierarchy_forecast_level:** The default level of the hierarchy to produce prediction/forecast on.\n",
+    "* **allocation_method:** \\[Optional] The disaggregation method to use if the hierarchy forecast level specified is below the define hierarchy training level. <br><i>(average historical proportions) 'average_historical_proportions'</i><br><i>(proportions of the historical averages) 'proportions_of_historical_average'</i>\n",
+    "\n",
+    "#### get_many_models_batch_inference_steps arguments\n",
+    "* **experiment:** The experiment used for inference run.\n",
+    "* **inference_data:** The data to use for inferencing. It should be the same schema as used for training.\n",
+    "* **compute_target:** The compute target that runs the inference pipeline.\n",
+    "* **node_count:** The number of compute nodes to be used for running the user script. We recommend to start with the number of cores per node (varies by compute sku).\n",
+    "* **process_count_per_node:** The number of processes per node.\n",
+    "* **train_run_id:** \\[Optional] The run id of the hierarchy training, by default it is the latest successful training hts run in the experiment.\n",
+    "* **train_experiment_name:** \\[Optional] The train experiment that contains the train pipeline. This one is only needed when the train pipeline is not in the same experiement as the inference pipeline.\n",
+    "* **process_count_per_node:** \\[Optional] The number of processes per node, by default it's 4."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.train.automl.runtime._hts.hts_parameters import HTSInferenceParameters\n",
+    "\n",
+    "inference_parameters = HTSInferenceParameters(\n",
+    "    hierarchy_forecast_level=\"store_id\",  # The setting is specific to this dataset and should be changed based on your dataset.\n",
+    "    allocation_method=\"proportions_of_historical_average\",\n",
+    ")\n",
+    "\n",
+    "steps = AutoMLPipelineBuilder.get_many_models_batch_inference_steps(\n",
+    "    experiment=experiment,\n",
+    "    inference_data=registered_inference,\n",
+    "    compute_target=compute_target,\n",
+    "    inference_pipeline_parameters=inference_parameters,\n",
+    "    node_count=2,\n",
+    "    process_count_per_node=8,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.pipeline.core import Pipeline\n",
+    "\n",
+    "inference_pipeline = Pipeline(ws, steps=steps)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "inference_run = experiment.submit(inference_pipeline)\n",
+    "inference_run.wait_for_completion(show_output=True)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Retrieve results\n",
+    "\n",
+    "Forecast results can be retrieved through the following code. The prediction results summary and the actual predictions are downloaded in forecast_results folder"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "forecasts = inference_run.get_pipeline_output(\"forecasts\")\n",
+    "forecasts.download(\"forecast_results\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Resbumit the Pipeline\n",
+    "\n",
+    "The inference pipeline can be submitted with different configurations."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "inference_run = experiment.submit(\n",
+    "    inference_pipeline, pipeline_parameters={\"hierarchy_forecast_level\": \"state\"}\n",
+    ")\n",
+    "inference_run.wait_for_completion(show_output=False)"
+   ]
+  }
+ ],
+ "metadata": {
+  "authors": [
+   {
+    "name": "jialiu"
+   }
+  ],
+  "categories": [
+   "how-to-use-azureml",
+   "automated-machine-learning"
+  ],
+  "kernelspec": {
+   "display_name": "Python 3.6 - AzureML",
+   "language": "python",
+   "name": "python3-azureml"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.6.8"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}

how-to-use-azureml/automated-machine-learning/forecasting-hierarchical-timeseries/auto-ml-forecasting-hierarchical-timeseries.yml

@@ -0,0 +1,4 @@
+name: auto-ml-forecasting-hierarchical-timeseries
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/forecasting-hierarchical-timeseries/update_env.yml

@@ -0,0 +1,3 @@
+dependencies:
+- pip:
+  - azureml-contrib-automl-pipeline-steps

how-to-use-azureml/automated-machine-learning/forecasting-many-models/README.md

@@ -0,0 +1,122 @@
+---
+page_type: sample
+languages:
+- python
+products:
+- azure-machine-learning
+description: Tutorial showing how to solve a complex machine learning time series forecasting problems at scale by using Azure Automated ML and Many Models solution accelerator.
+---
+
+![Many Models Solution Accelerator Banner](images/mmsa.png)
+# Many Models Solution Accelerator
+
+<!-- 
+Guidelines on README format: https://review.docs.microsoft.com/help/onboard/admin/samples/concepts/readme-template?branch=master
+
+Guidance on onboarding samples to docs.microsoft.com/samples: https://review.docs.microsoft.com/help/onboard/admin/samples/process/onboarding?branch=master
+
+Taxonomies for products and languages: https://review.docs.microsoft.com/new-hope/information-architecture/metadata/taxonomies?branch=master
+-->
+
+In the real world, many problems can be too complex to be solved by a single machine learning model. Whether that be predicting sales for each individual store, building a predictive maintanence model for hundreds of oil wells, or tailoring an experience to individual users, building a model for each instance can lead to improved results on many machine learning problems.
+
+This Pattern is very common across a wide variety of industries and applicable to many real world use cases. Below are some examples we have seen where this pattern is being used.
+
+- Energy and utility companies building predictive maintenance models for thousands of oil wells, hundreds of wind turbines or hundreds of smart meters
+
+- Retail organizations building workforce optimization models for thousands of stores, campaign promotion propensity models, Price optimization models for hundreds of thousands of products they sell
+
+- Restaurant chains building demand forecasting models across thousands of restaurants  
+
+- Banks and financial institutes building models for cash replenishment for ATM Machine and for several ATMs or building personalized models for individuals
+
+- Enterprises building revenue forecasting models at each division level
+
+- Document management companies building text analytics and legal document search models per each state
+
+Azure Machine Learning (AML) makes it easy to train, operate, and manage hundreds or even thousands of models. This repo will walk you through the end to end process of creating a many models solution from training to scoring to monitoring.
+
+## Prerequisites
+
+To use this solution accelerator, all you need is access to an [Azure subscription](https://azure.microsoft.com/free/) and an [Azure Machine Learning Workspace](https://docs.microsoft.com/azure/machine-learning/how-to-manage-workspace) that you'll create below.
+
+While it's not required, a basic understanding of Azure Machine Learning will be helpful for understanding the solution. The following resources can help introduce you to AML:
+
+1. [Azure Machine Learning Overview](https://azure.microsoft.com/services/machine-learning/)
+2. [Azure Machine Learning Tutorials](https://docs.microsoft.com/azure/machine-learning/tutorial-1st-experiment-sdk-setup)
+3. [Azure Machine Learning Sample Notebooks on Github](https://github.com/Azure/azureml-examples)
+
+## Getting started
+
+### 1. Deploy Resources
+
+Start by deploying the resources to Azure. The button below will deploy Azure Machine Learning and its related resources:
+
+<a href="https://portal.azure.com/#create/Microsoft.Template/uri/https%3A%2F%2Fraw.githubusercontent.com%2Fmicrosoft%2Fsolution-accelerator-many-models%2Fmaster%2Fazuredeploy.json" target="_blank">
+    <img src="http://azuredeploy.net/deploybutton.png"/>
+</a>
+
+### 2. Configure Development Environment
+
+Next you'll need to configure your [development environment](https://docs.microsoft.com/azure/machine-learning/how-to-configure-environment) for Azure Machine Learning. We recommend using a [Compute Instance](https://docs.microsoft.com/azure/machine-learning/how-to-configure-environment#compute-instance) as it's the fastest way to get up and running.
+
+### 3. Run Notebooks
+
+Once your development environment is set up, run through the Jupyter Notebooks sequentially following the steps outlined. By the end, you'll know how to train, score, and make predictions using the many models pattern on Azure Machine Learning.
+
+![Sequence of Notebooks](./images/mmsa-overview.png)
+
+
+## Contents
+
+In this repo, you'll train and score a forecasting model for each orange juice brand and for each store at a (simulated) grocery chain. By the end, you'll have forecasted sales by using up to 11,973 models to predict sales for the next few weeks.
+
+The data used in this sample is simulated based on the [Dominick's Orange Juice Dataset](http://www.cs.unitn.it/~taufer/QMMA/L10-OJ-Data.html#(1)), sales data from a Chicago area grocery store.
+
+<img src="images/Flow_map.png" width="1000">
+
+### Using Automated ML to train the models:
+
+The [`auto-ml-forecasting-many-models.ipynb`](./auto-ml-forecasting-many-models.ipynb) noteboook is a guided solution accelerator that demonstrates steps from data preparation, to model training, and forecasting on train models as well as operationalizing the solution.
+
+## How-to-videos
+
+Watch these how-to-videos for a step by step walk-through of the many model solution accelerator to learn how to setup your models using Automated ML.
+
+### Automated ML
+
+[![Watch the video](https://media.giphy.com/media/dWUKfameudyNGRnp1t/giphy.gif)](https://channel9.msdn.com/Shows/Docs-AI/Building-Large-Scale-Machine-Learning-Forecasting-Models-using-Azure-Machine-Learnings-Automated-ML)
+
+## Key concepts
+
+### ParallelRunStep
+
+[ParallelRunStep](https://docs.microsoft.com/en-us/python/api/azureml-pipeline-steps/azureml.pipeline.steps.parallel_run_step.parallelrunstep?view=azure-ml-py) enables the parallel training of models and is commonly used for batch inferencing. This [document](https://docs.microsoft.com/azure/machine-learning/how-to-use-parallel-run-step) walks through some of the key concepts around ParallelRunStep.
+
+### Pipelines
+
+[Pipelines](https://docs.microsoft.com/azure/machine-learning/concept-ml-pipelines) allow you to create workflows in your machine learning projects. These workflows have a number of benefits including speed, simplicity, repeatability, and modularity.
+
+### Automated Machine Learning
+
+[Automated Machine Learning](https://docs.microsoft.com/azure/machine-learning/concept-automated-ml) also referred to as automated ML or AutoML, is the process of automating the time consuming, iterative tasks of machine learning model development. It allows data scientists, analysts, and developers to build ML models with high scale, efficiency, and productivity all while sustaining model quality.
+
+### Other Concepts
+
+In additional to ParallelRunStep, Pipelines and Automated Machine Learning, you'll also be working with the following concepts including [workspace](https://docs.microsoft.com/azure/machine-learning/concept-workspace), [datasets](https://docs.microsoft.com/azure/machine-learning/concept-data#datasets), [compute targets](https://docs.microsoft.com/azure/machine-learning/concept-compute-target#train), [python script steps](https://docs.microsoft.com/python/api/azureml-pipeline-steps/azureml.pipeline.steps.python_script_step.pythonscriptstep?view=azure-ml-py), and [Azure Open Datasets](https://azure.microsoft.com/services/open-datasets/).
+
+## Contributing
+
+This project welcomes contributions and suggestions. To learn more visit the [contributing](../../../CONTRIBUTING.md) section.
+
+Most contributions require you to agree to a Contributor License Agreement (CLA)
+declaring that you have the right to, and actually do, grant us
+the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.
+
+When you submit a pull request, a CLA bot will automatically determine whether you need to provide
+a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions
+provided by the bot. You will only need to do this once across all repos using our CLA.
+
+This project has adopted the [Microsoft Open Source Code of Conduct](https://opensource.microsoft.com/codeofconduct/).
+For more information see the [Code of Conduct FAQ](https://opensource.microsoft.com/codeofconduct/faq/) or
+contact [opencode@microsoft.com](mailto:opencode@microsoft.com) with any additional questions or comments.

how-to-use-azureml/automated-machine-learning/forecasting-many-models/auto-ml-forecasting-many-models.ipynb

@@ -0,0 +1,746 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Copyright (c) Microsoft Corporation. All rights reserved.\n",
+    "\n",
+    "Licensed under the MIT License."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "![Impressions](https://PixelServer20190423114238.azurewebsites.net/api/impressions/MachineLearningNotebooks/how-to-use-azureml/automated-machine-learning/forecasting-hierarchical-timeseries/auto-ml-forecasting-hierarchical-timeseries.png)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Many Models - Automated ML\n",
+    "**_Generate many models time series forecasts with Automated Machine Learning_**\n",
+    "\n",
+    "---"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "For this notebook we are using a synthetic dataset portraying sales data to predict the the quantity of a vartiety of product skus across several states, stores, and product categories.\n",
+    "\n",
+    "**NOTE: There are limits on how many runs we can do in parallel per workspace, and we currently recommend to set the parallelism to maximum of 320 runs per experiment per workspace. If users want to have more parallelism and increase this limit they might encounter Too Many Requests errors (HTTP 429).**"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Prerequisites\n",
+    "You'll need to create a compute Instance by following the instructions in the [EnvironmentSetup.md](../Setup_Resources/EnvironmentSetup.md)."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 1.0 Set up workspace, datastore, experiment"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613003526897
+    }
+   },
+   "outputs": [],
+   "source": [
+    "import azureml.core\n",
+    "from azureml.core import Workspace, Datastore\n",
+    "import pandas as pd\n",
+    "\n",
+    "# Set up your workspace\n",
+    "ws = Workspace.from_config()\n",
+    "ws.get_details()\n",
+    "\n",
+    "# Set up your datastores\n",
+    "dstore = ws.get_default_datastore()\n",
+    "\n",
+    "output = {}\n",
+    "output[\"SDK version\"] = azureml.core.VERSION\n",
+    "output[\"Subscription ID\"] = ws.subscription_id\n",
+    "output[\"Workspace\"] = ws.name\n",
+    "output[\"Resource Group\"] = ws.resource_group\n",
+    "output[\"Location\"] = ws.location\n",
+    "output[\"Default datastore name\"] = dstore.name\n",
+    "pd.set_option(\"display.max_colwidth\", -1)\n",
+    "outputDf = pd.DataFrame(data=output, index=[\"\"])\n",
+    "outputDf.T"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Choose an experiment"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613003540729
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.core import Experiment\n",
+    "\n",
+    "experiment = Experiment(ws, \"automl-many-models\")\n",
+    "\n",
+    "print(\"Experiment name: \" + experiment.name)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 2.0 Data\n",
+    "\n",
+    "This notebook uses simulated orange juice sales data to walk you through the process of training many models on Azure Machine Learning using Automated ML. \n",
+    "\n",
+    "The time series data used in this example was simulated based on the University of Chicago's Dominick's Finer Foods dataset which featured two years of sales of 3 different orange juice brands for individual stores. The full simulated dataset includes 3,991 stores with 3 orange juice brands each thus allowing 11,973 models to be trained to showcase the power of the many models pattern.\n",
+    "\n",
+    "  \n",
+    "In this notebook, two datasets will be created: one with all 11,973 files and one with only 10 files that can be used to quickly test and debug. For each dataset, you'll be walked through the process of:\n",
+    "\n",
+    "1. Registering the blob container as a Datastore to the Workspace\n",
+    "2. Registering a tabular dataset to the Workspace"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "nteract": {
+     "transient": {
+      "deleting": false
+     }
+    }
+   },
+   "source": [
+    "### 2.1 Data Preparation\n",
+    "The OJ data is available in the public blob container. The data is split to be used for training and for inferencing. For the current dataset, the data was split on time column ('WeekStarting') before and after '1992-5-28' .\n",
+    "\n",
+    "The container has\n",
+    "<ol>\n",
+    "    <li><b>'oj-data-tabular'</b> and <b>'oj-inference-tabular'</b> folders that contains training and inference data respectively for the 11,973 models. </li>\n",
+    "    <li>It also has <b>'oj-data-small-tabular'</b> and <b>'oj-inference-small-tabular'</b> folders that has training and inference data for 10 models.</li>\n",
+    "</ol>\n",
+    "\n",
+    "To create the [TabularDataset](https://docs.microsoft.com/en-us/python/api/azureml-core/azureml.data.tabular_dataset.tabulardataset?view=azure-ml-py) needed for the ParallelRunStep, you first need to register the blob container to the workspace."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "nteract": {
+     "transient": {
+      "deleting": false
+     }
+    }
+   },
+   "source": [
+    "<b> To use your own data, put your own data in a blobstore folder. As shown it can be one file or multiple files. We can then register datastore using that blob as shown below.\n",
+    "    \n",
+    "<h3> How sample data in blob store looks like</h3>\n",
+    "\n",
+    "['oj-data-tabular'](https://ms.portal.azure.com/#blade/Microsoft_Azure_Storage/ContainerMenuBlade/overview/storageAccountId/%2Fsubscriptions%2F102a16c3-37d3-48a8-9237-4c9b1e8e80e0%2FresourceGroups%2FAutoMLSampleNotebooksData%2Fproviders%2FMicrosoft.Storage%2FstorageAccounts%2Fautomlsamplenotebookdata/path/automl-sample-notebook-data/etag/%220x8D84EAA65DE50B7%22/defaultEncryptionScope/%24account-encryption-key/denyEncryptionScopeOverride//defaultId//publicAccessVal/Container)</b>\n",
+    "![image-4.png](mm-1.png)\n",
+    "\n",
+    "['oj-inference-tabular'](https://ms.portal.azure.com/#blade/Microsoft_Azure_Storage/ContainerMenuBlade/overview/storageAccountId/%2Fsubscriptions%2F102a16c3-37d3-48a8-9237-4c9b1e8e80e0%2FresourceGroups%2FAutoMLSampleNotebooksData%2Fproviders%2FMicrosoft.Storage%2FstorageAccounts%2Fautomlsamplenotebookdata/path/automl-sample-notebook-data/etag/%220x8D84EAA65DE50B7%22/defaultEncryptionScope/%24account-encryption-key/denyEncryptionScopeOverride//defaultId//publicAccessVal/Container)\n",
+    "![image-3.png](mm-2.png)\n",
+    "\n",
+    "['oj-data-small-tabular'](https://ms.portal.azure.com/#blade/Microsoft_Azure_Storage/ContainerMenuBlade/overview/storageAccountId/%2Fsubscriptions%2F102a16c3-37d3-48a8-9237-4c9b1e8e80e0%2FresourceGroups%2FAutoMLSampleNotebooksData%2Fproviders%2FMicrosoft.Storage%2FstorageAccounts%2Fautomlsamplenotebookdata/path/automl-sample-notebook-data/etag/%220x8D84EAA65DE50B7%22/defaultEncryptionScope/%24account-encryption-key/denyEncryptionScopeOverride//defaultId//publicAccessVal/Container)\n",
+    "\n",
+    "![image-5.png](mm-3.png)\n",
+    "\n",
+    "['oj-inference-small-tabular'](https://ms.portal.azure.com/#blade/Microsoft_Azure_Storage/ContainerMenuBlade/overview/storageAccountId/%2Fsubscriptions%2F102a16c3-37d3-48a8-9237-4c9b1e8e80e0%2FresourceGroups%2FAutoMLSampleNotebooksData%2Fproviders%2FMicrosoft.Storage%2FstorageAccounts%2Fautomlsamplenotebookdata/path/automl-sample-notebook-data/etag/%220x8D84EAA65DE50B7%22/defaultEncryptionScope/%24account-encryption-key/denyEncryptionScopeOverride//defaultId//publicAccessVal/Container)\n",
+    "![image-6.png](mm-4.png)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### 2.2 Register the blob container as DataStore\n",
+    "\n",
+    "A Datastore is a place where data can be stored that is then made accessible to a compute either by means of mounting or copying the data to the compute target.\n",
+    "\n",
+    "Please refer to [Datastore](https://docs.microsoft.com/en-us/python/api/azureml-core/azureml.core.datastore(class)?view=azure-ml-py) documentation on how to access data from Datastore.\n",
+    "\n",
+    "In this next step, we will be registering blob storage as datastore to the Workspace."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.core import Datastore\n",
+    "\n",
+    "# Please change the following to point to your own blob container and pass in account_key\n",
+    "blob_datastore_name = \"automl_many_models\"\n",
+    "container_name = \"automl-sample-notebook-data\"\n",
+    "account_name = \"automlsamplenotebookdata\"\n",
+    "\n",
+    "oj_datastore = Datastore.register_azure_blob_container(\n",
+    "    workspace=ws,\n",
+    "    datastore_name=blob_datastore_name,\n",
+    "    container_name=container_name,\n",
+    "    account_name=account_name,\n",
+    "    create_if_not_exists=True,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### 2.3 Using tabular datasets \n",
+    "\n",
+    "Now that the datastore is available from the Workspace, [TabularDataset](https://docs.microsoft.com/en-us/python/api/azureml-core/azureml.data.tabular_dataset.tabulardataset?view=azure-ml-py) can be created. Datasets in Azure Machine Learning are references to specific data in a Datastore.  We are using TabularDataset, so that users who have their data which can be in one or many files (*.parquet or *.csv) and have not split up data according to group columns needed for training, can do so using out of box support for 'partiion_by' feature of TabularDataset shown in section 5.0 below."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613007017296
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.core import Dataset\n",
+    "\n",
+    "ds_name_small = \"oj-data-small-tabular\"\n",
+    "input_ds_small = Dataset.Tabular.from_delimited_files(\n",
+    "    path=oj_datastore.path(ds_name_small + \"/\"), validate=False\n",
+    ")\n",
+    "\n",
+    "inference_name_small = \"oj-inference-small-tabular\"\n",
+    "inference_ds_small = Dataset.Tabular.from_delimited_files(\n",
+    "    path=oj_datastore.path(inference_name_small + \"/\"), validate=False\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 3.0 Build the training pipeline\n",
+    "Now that the dataset, WorkSpace, and datastore are set up, we can put together a pipeline for training.\n",
+    "\n",
+    "> Note that if you have an AzureML Data Scientist role, you will not have permission to create compute resources. Talk to your workspace or IT admin to create the compute targets described in this section, if they do not already exist."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Choose a compute target\n",
+    "\n",
+    "You will need to create a [compute target](https://docs.microsoft.com/en-us/azure/machine-learning/how-to-set-up-training-targets#amlcompute) for your AutoML run. In this tutorial, you create AmlCompute as your training compute resource.\n",
+    "\n",
+    "\\*\\*Creation of AmlCompute takes approximately 5 minutes.**\n",
+    "\n",
+    "If the AmlCompute with that name is already in your workspace this code will skip the creation process. As with other Azure services, there are limits on certain resources (e.g. AmlCompute) associated with the Azure Machine Learning service. Please read this [article](https://docs.microsoft.com/en-us/azure/machine-learning/how-to-manage-quotas) on the default limits and how to request more quota."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613007037308
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.core.compute import ComputeTarget, AmlCompute\n",
+    "\n",
+    "# Name your cluster\n",
+    "compute_name = \"mm-compute\"\n",
+    "\n",
+    "\n",
+    "if compute_name in ws.compute_targets:\n",
+    "    compute_target = ws.compute_targets[compute_name]\n",
+    "    if compute_target and type(compute_target) is AmlCompute:\n",
+    "        print(\"Found compute target: \" + compute_name)\n",
+    "else:\n",
+    "    print(\"Creating a new compute target...\")\n",
+    "    provisioning_config = AmlCompute.provisioning_configuration(\n",
+    "        vm_size=\"STANDARD_D16S_V3\", max_nodes=20\n",
+    "    )\n",
+    "    # Create the compute target\n",
+    "    compute_target = ComputeTarget.create(ws, compute_name, provisioning_config)\n",
+    "\n",
+    "    # Can poll for a minimum number of nodes and for a specific timeout.\n",
+    "    # If no min node count is provided it will use the scale settings for the cluster\n",
+    "    compute_target.wait_for_completion(\n",
+    "        show_output=True, min_node_count=None, timeout_in_minutes=20\n",
+    "    )\n",
+    "\n",
+    "    # For a more detailed view of current cluster status, use the 'status' property\n",
+    "    print(compute_target.status.serialize())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Set up training parameters\n",
+    "\n",
+    "This dictionary defines the AutoML and many models settings. For this forecasting task we need to define several settings inncluding the name of the time column, the maximum forecast horizon, and the partition column name definition.\n",
+    "\n",
+    "| Property                           | Description|\n",
+    "| :---------------                   | :------------------- |\n",
+    "| **task**                           | forecasting |\n",
+    "| **primary_metric**                 | This is the metric that you want to optimize.<br> Forecasting supports the following primary metrics <br><i>spearman_correlation</i><br><i>normalized_root_mean_squared_error</i><br><i>r2_score</i><br><i>normalized_mean_absolute_error</i> |\n",
+    "| **blocked_models**                 | Blocked models won't be used by AutoML. |\n",
+    "| **iteration_timeout_minutes**      | Maximum amount of time in minutes that the model can train. This is optional but provides customers with greater control on exit criteria. |\n",
+    "| **iterations**                     | Number of models to train. This is optional but provides customers with greater control on exit criteria. |\n",
+    "| **experiment_timeout_hours**       | Maximum amount of time in hours that the experiment can take before it terminates. This is optional but provides customers with greater control on exit criteria. |\n",
+    "| **label_column_name**              | The name of the label column. |\n",
+    "| **forecast_horizon**               | The forecast horizon is how many periods forward you would like to forecast. This integer horizon is in units of the timeseries frequency (e.g. daily, weekly). Periods are inferred from your data. |\n",
+    "| **n_cross_validations**            | Number of cross validation splits. Rolling Origin Validation is used to split time-series in a temporally consistent way. |\n",
+    "| **enable_early_stopping**          | Flag to enable early termination if the score is not improving in the short term. |\n",
+    "| **time_column_name**               | The name of your time column. |\n",
+    "| **enable_engineered_explanations** | Engineered feature explanations will be downloaded if enable_engineered_explanations flag is set to True. By default it is set to False to save storage space. |\n",
+    "| **time_series_id_column_name**     | The column names used to uniquely identify timeseries in data that has multiple rows with the same timestamp. |\n",
+    "| **track_child_runs**               | Flag to disable tracking of child runs. Only best run is tracked if the flag is set to False (this includes the model and metrics of the run). |\n",
+    "| **pipeline_fetch_max_batch_size**  | Determines how many pipelines (training algorithms) to fetch at a time for training, this helps reduce throttling when training at large scale. |\n",
+    "| **partition_column_names**         | The names of columns used to group your models. For timeseries, the groups must not split up individual time-series. That is, each group must contain one or more whole time-series. |"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "gather": {
+     "logged": 1613007061544
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from azureml.train.automl.runtime._many_models.many_models_parameters import (\n",
+    "    ManyModelsTrainParameters,\n",
+    ")\n",
+    "\n",
+    "partition_column_names = [\"Store\", \"Brand\"]\n",
+    "automl_settings = {\n",
+    "    \"task\": \"forecasting\",\n",
+    "    \"primary_metric\": \"normalized_root_mean_squared_error\",\n",
+    "    \"iteration_timeout_minutes\": 10,  # This needs to be changed based on the dataset. We ask customer to explore how long training is taking before settings this value\n",
+    "    \"iterations\": 15,\n",
+    "    \"experiment_timeout_hours\": 0.25,\n",
+    "    \"label_column_name\": \"Quantity\",\n",
+    "    \"n_cross_validations\": 3,\n",
+    "    \"time_column_name\": \"WeekStarting\",\n",
+    "    \"drop_column_names\": \"Revenue\",\n",
+    "    \"max_horizon\": 6,\n",
+    "    \"grain_column_names\": partition_column_names,\n",
+    "    \"track_child_runs\": False,\n",
+    "}\n",
+    "\n",
+    "mm_paramters = ManyModelsTrainParameters(\n",
+    "    automl_settings=automl_settings, partition_column_names=partition_column_names\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Set up many models pipeline"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Parallel run step is leveraged to train multiple models at once. To configure the ParallelRunConfig you will need to determine the appropriate number of workers and nodes for your use case. The process_count_per_node is based off the number of cores of the compute VM. The node_count will determine the number of master nodes to use, increasing the node count will speed up the training process.\n",
+    "\n",
+    "| Property                           | Description|\n",
+    "| :---------------                   | :------------------- |\n",
+    "| **experiment**                     | The experiment used for training. |\n",
+    "| **train_data**                     | The file dataset to be used as input to the training run. |\n",
+    "| **node_count**                     | The number of compute nodes to be used for running the user script. We recommend to start with 3 and increase the node_count if the training time is taking too long. |\n",
+    "| **process_count_per_node**         | Process count per node, we recommend 2:1 ratio for number of cores: number of processes per node. eg. If node has 16 cores then configure 8 or less process count per node or optimal performance. |\n",
+    "| **train_pipeline_parameters**      | The set of configuration parameters defined in the previous section. |\n",
+    "\n",
+    "Calling this method will create a new aggregated dataset which is generated dynamically on pipeline execution."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.contrib.automl.pipeline.steps import AutoMLPipelineBuilder\n",
+    "\n",
+    "\n",
+    "training_pipeline_steps = AutoMLPipelineBuilder.get_many_models_train_steps(\n",
+    "    experiment=experiment,\n",
+    "    train_data=input_ds_small,\n",
+    "    compute_target=compute_target,\n",
+    "    node_count=2,\n",
+    "    process_count_per_node=8,\n",
+    "    run_invocation_timeout=920,\n",
+    "    train_pipeline_parameters=mm_paramters,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.pipeline.core import Pipeline\n",
+    "\n",
+    "training_pipeline = Pipeline(ws, steps=training_pipeline_steps)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Submit the pipeline to run\n",
+    "Next we submit our pipeline to run. The whole training pipeline takes about 40m using a STANDARD_D16S_V3 VM with our current ParallelRunConfig setting."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "training_run = experiment.submit(training_pipeline)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "training_run.wait_for_completion(show_output=False)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Check the run status, if training_run is in completed state, continue to forecasting. If training_run is in another state, check the portal for failures."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 5.0 Publish and schedule the train pipeline (Optional)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### 5.1 Publish the pipeline\n",
+    "\n",
+    "Once you have a pipeline you're happy with, you can publish a pipeline so you can call it programmatically later on. See this [tutorial](https://docs.microsoft.com/en-us/azure/machine-learning/how-to-create-your-first-pipeline#publish-a-pipeline) for additional information on publishing and calling pipelines."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# published_pipeline = training_pipeline.publish(name = 'automl_train_many_models',\n",
+    "#                                                description = 'train many models',\n",
+    "#                                                version = '1',\n",
+    "#                                                continue_on_step_failure = False)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### 7.2 Schedule the pipeline\n",
+    "You can also [schedule the pipeline](https://docs.microsoft.com/en-us/azure/machine-learning/how-to-schedule-pipelines) to run on a time-based or change-based schedule. This could be used to automatically retrain models every month or based on another trigger such as data drift."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# from azureml.pipeline.core import Schedule, ScheduleRecurrence\n",
+    "\n",
+    "# training_pipeline_id = published_pipeline.id\n",
+    "\n",
+    "# recurrence = ScheduleRecurrence(frequency=\"Month\", interval=1, start_time=\"2020-01-01T09:00:00\")\n",
+    "# recurring_schedule = Schedule.create(ws, name=\"automl_training_recurring_schedule\",\n",
+    "#                                      description=\"Schedule Training Pipeline to run on the first day of every month\",\n",
+    "#                                      pipeline_id=training_pipeline_id,\n",
+    "#                                      experiment_name=experiment.name,\n",
+    "#                                      recurrence=recurrence)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 6.0 Forecasting"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Set up output dataset for inference data\n",
+    "Output of inference can be represented as [OutputFileDatasetConfig](https://docs.microsoft.com/en-us/python/api/azureml-core/azureml.data.output_dataset_config.outputdatasetconfig?view=azure-ml-py) object and OutputFileDatasetConfig can be registered as a dataset. "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.data import OutputFileDatasetConfig\n",
+    "\n",
+    "output_inference_data_ds = OutputFileDatasetConfig(\n",
+    "    name=\"many_models_inference_output\", destination=(dstore, \"oj/inference_data/\")\n",
+    ").register_on_complete(name=\"oj_inference_data_ds\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "For many models we need to provide the ManyModelsInferenceParameters object.\n",
+    "\n",
+    "#### ManyModelsInferenceParameters arguments\n",
+    "| Property                           | Description|\n",
+    "| :---------------                   | :------------------- |\n",
+    "| **partition_column_names**         | List of column names that identifies groups.                                           |\n",
+    "| **target_column_name**             | \\[Optional] Column name only if the inference dataset has the target. |\n",
+    "| **time_column_name**               | \\[Optional] Column name only if it is timeseries. |\n",
+    "| **many_models_run_id**             | \\[Optional] Many models run id where models were trained. |\n",
+    "\n",
+    "#### get_many_models_batch_inference_steps arguments\n",
+    "| Property                           | Description|\n",
+    "| :---------------                   | :------------------- |\n",
+    "| **experiment**                     | The experiment used for inference run. |\n",
+    "| **inference_data**                 | The data to use for inferencing. It should be the same schema as used for training.\n",
+    "| **compute_target** The compute target that runs the inference pipeline.|\n",
+    "| **node_count**                     | The number of compute nodes to be used for running the user script. We recommend to start with the number of cores per node (varies by compute sku). |\n",
+    "| **process_count_per_node** The number of processes per node.\n",
+    "| **train_run_id**                   | \\[Optional] The run id of the hierarchy training, by default it is the latest successful training many model run in the experiment. |\n",
+    "| **train_experiment_name**          | \\[Optional] The train experiment that contains the train pipeline. This one is only needed when the train pipeline is not in the same experiement as the inference pipeline. |\n",
+    "| **process_count_per_node**         | \\[Optional] The number of processes per node, by default it's 4. |"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.contrib.automl.pipeline.steps import AutoMLPipelineBuilder\n",
+    "from azureml.train.automl.runtime._many_models.many_models_parameters import (\n",
+    "    ManyModelsInferenceParameters,\n",
+    ")\n",
+    "\n",
+    "mm_parameters = ManyModelsInferenceParameters(\n",
+    "    partition_column_names=[\"Store\", \"Brand\"],\n",
+    "    time_column_name=\"WeekStarting\",\n",
+    "    target_column_name=\"Quantity\",\n",
+    ")\n",
+    "\n",
+    "inference_steps = AutoMLPipelineBuilder.get_many_models_batch_inference_steps(\n",
+    "    experiment=experiment,\n",
+    "    inference_data=inference_ds_small,\n",
+    "    node_count=2,\n",
+    "    process_count_per_node=8,\n",
+    "    compute_target=compute_target,\n",
+    "    run_invocation_timeout=300,\n",
+    "    output_datastore=output_inference_data_ds,\n",
+    "    train_run_id=training_run.id,\n",
+    "    train_experiment_name=training_run.experiment.name,\n",
+    "    inference_pipeline_parameters=mm_parameters,\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.pipeline.core import Pipeline\n",
+    "\n",
+    "inference_pipeline = Pipeline(ws, steps=inference_steps)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "inference_run = experiment.submit(inference_pipeline)\n",
+    "inference_run.wait_for_completion(show_output=False)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Retrieve results\n",
+    "\n",
+    "The forecasting pipeline forecasts the orange juice quantity for a Store by Brand. The pipeline returns one file with the predictions for each store and outputs the result to the forecasting_output Blob container. The details of the blob container is listed in 'forecasting_output.txt' under Outputs+logs. \n",
+    "\n",
+    "The following code snippet:\n",
+    "1. Downloads the contents of the output folder that is passed in the parallel run step \n",
+    "2. Reads the parallel_run_step.txt file that has the predictions as pandas dataframe and \n",
+    "3. Displays the top 10 rows of the predictions"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from azureml.contrib.automl.pipeline.steps.utilities import get_output_from_mm_pipeline\n",
+    "\n",
+    "forecasting_results_name = \"forecasting_results\"\n",
+    "forecasting_output_name = \"many_models_inference_output\"\n",
+    "forecast_file = get_output_from_mm_pipeline(\n",
+    "    inference_run, forecasting_results_name, forecasting_output_name\n",
+    ")\n",
+    "df = pd.read_csv(forecast_file, delimiter=\" \", header=None)\n",
+    "df.columns = [\n",
+    "    \"Week Starting\",\n",
+    "    \"Store\",\n",
+    "    \"Brand\",\n",
+    "    \"Quantity\",\n",
+    "    \"Advert\",\n",
+    "    \"Price\",\n",
+    "    \"Revenue\",\n",
+    "    \"Predicted\",\n",
+    "]\n",
+    "print(\n",
+    "    \"Prediction has \", df.shape[0], \" rows. Here the first 10 rows are being displayed.\"\n",
+    ")\n",
+    "df.head(10)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 7.0 Publish and schedule the inference pipeline (Optional)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### 7.1 Publish the pipeline\n",
+    "\n",
+    "Once you have a pipeline you're happy with, you can publish a pipeline so you can call it programmatically later on. See this [tutorial](https://docs.microsoft.com/en-us/azure/machine-learning/how-to-create-your-first-pipeline#publish-a-pipeline) for additional information on publishing and calling pipelines."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# published_pipeline_inf = inference_pipeline.publish(name = 'automl_forecast_many_models',\n",
+    "#                                                     description = 'forecast many models',\n",
+    "#                                                     version = '1',\n",
+    "#                                                     continue_on_step_failure = False)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### 7.2 Schedule the pipeline\n",
+    "You can also [schedule the pipeline](https://docs.microsoft.com/en-us/azure/machine-learning/how-to-schedule-pipelines) to run on a time-based or change-based schedule. This could be used to automatically retrain or forecast models every month or based on another trigger such as data drift."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# from azureml.pipeline.core import Schedule, ScheduleRecurrence\n",
+    "\n",
+    "# forecasting_pipeline_id = published_pipeline.id\n",
+    "\n",
+    "# recurrence = ScheduleRecurrence(frequency=\"Month\", interval=1, start_time=\"2020-01-01T09:00:00\")\n",
+    "# recurring_schedule = Schedule.create(ws, name=\"automl_forecasting_recurring_schedule\",\n",
+    "#                             description=\"Schedule Forecasting Pipeline to run on the first day of every week\",\n",
+    "#                             pipeline_id=forecasting_pipeline_id,\n",
+    "#                             experiment_name=experiment.name,\n",
+    "#                             recurrence=recurrence)"
+   ]
+  }
+ ],
+ "metadata": {
+  "authors": [
+   {
+    "name": "jialiu"
+   }
+  ],
+  "categories": [
+   "how-to-use-azureml",
+   "automated-machine-learning"
+  ],
+  "kernelspec": {
+   "display_name": "Python 3.6 - AzureML",
+   "language": "python",
+   "name": "python3-azureml"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.6.8"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}

how-to-use-azureml/automated-machine-learning/forecasting-many-models/auto-ml-forecasting-many-models.yml

@@ -0,0 +1,4 @@
+name: auto-ml-forecasting-many-models
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/forecasting-many-models/update_env.yml

@@ -0,0 +1,3 @@
+dependencies:
+- pip:
+  - azureml-contrib-automl-pipeline-steps

how-to-use-azureml/automated-machine-learning/forecasting-orange-juice-sales/auto-ml-forecasting-orange-juice-sales.yml

@@ -0,0 +1,4 @@
+name: auto-ml-forecasting-orange-juice-sales
+dependencies:
+- pip:
+  - azureml-sdk

how-to-use-azureml/automated-machine-learning/forecasting-orange-juice-sales/forecasting_script.py

@@ -0,0 +1,61 @@
+"""
+This is the script that is executed on the compute instance. It relies
+on the model.pkl file which is uploaded along with this script to the
+compute instance.
+"""
+
+import argparse
+from azureml.core import Dataset, Run
+from sklearn.externals import joblib
+from pandas.tseries.frequencies import to_offset
+
+parser = argparse.ArgumentParser()
+parser.add_argument(
+    "--target_column_name",
+    type=str,
+    dest="target_column_name",
+    help="Target Column Name",
+)
+parser.add_argument(
+    "--test_dataset", type=str, dest="test_dataset", help="Test Dataset"
+)
+
+args = parser.parse_args()
+target_column_name = args.target_column_name
+test_dataset_id = args.test_dataset
+
+run = Run.get_context()
+ws = run.experiment.workspace
+
+# get the input dataset by id
+test_dataset = Dataset.get_by_id(ws, id=test_dataset_id)
+
+X_test = test_dataset.to_pandas_dataframe().reset_index(drop=True)
+y_test = X_test.pop(target_column_name).values
+
+# generate forecast
+fitted_model = joblib.load("model.pkl")
+# We have default quantiles values set as below(95th percentile)
+quantiles = [0.025, 0.5, 0.975]
+predicted_column_name = "predicted"
+PI = "prediction_interval"
+fitted_model.quantiles = quantiles
+pred_quantiles = fitted_model.forecast_quantiles(X_test)
+pred_quantiles[PI] = pred_quantiles[[min(quantiles), max(quantiles)]].apply(
+    lambda x: "[{}, {}]".format(x[0], x[1]), axis=1
+)
+X_test[target_column_name] = y_test
+X_test[PI] = pred_quantiles[PI]
+X_test[predicted_column_name] = pred_quantiles[0.5]
+# drop rows where prediction or actuals are nan
+# happens because of missing actuals
+# or at edges of time due to lags/rolling windows
+clean = X_test[
+    X_test[[target_column_name, predicted_column_name]].notnull().all(axis=1)
+]
+
+file_name = "outputs/predictions.csv"
+export_csv = clean.to_csv(file_name, header=True, index=False)  # added Index
+
+# Upload the predictions into artifacts
+run.upload_file(name=file_name, path_or_stream=file_name)

how-to-use-azureml/automated-machine-learning/forecasting-orange-juice-sales/run_forecast.py

@@ -0,0 +1,49 @@
+import os
+import shutil
+from azureml.core import ScriptRunConfig
+
+
+def run_remote_inference(
+    test_experiment,
+    compute_target,
+    train_run,
+    test_dataset,
+    target_column_name,
+    inference_folder="./forecast",
+):
+    # Create local directory to copy the model.pkl and forecsting_script.py files into.
+    # These files will be uploaded to and executed on the compute instance.
+    os.makedirs(inference_folder, exist_ok=True)
+    shutil.copy("forecasting_script.py", inference_folder)
+
+    train_run.download_file(
+        "outputs/model.pkl", os.path.join(inference_folder, "model.pkl")
+    )
+
+    inference_env = train_run.get_environment()
+
+    config = ScriptRunConfig(
+        source_directory=inference_folder,
+        script="forecasting_script.py",
+        arguments=[
+            "--target_column_name",
+            target_column_name,
+            "--test_dataset",
+            test_dataset.as_named_input(test_dataset.name),
+        ],
+        compute_target=compute_target,
+        environment=inference_env,
+    )
+
+    run = test_experiment.submit(
+        config,
+        tags={
+            "training_run_id": train_run.id,
+            "run_algorithm": train_run.properties["run_algorithm"],
+            "valid_score": train_run.properties["score"],
+            "primary_metric": train_run.properties["primary_metric"],
+        },
+    )
+
+    run.log("run_algorithm", run.tags["run_algorithm"])
+    return run