Tutorial: Consume Azure Machine Learning models in Power BI

This tutorial walks you through creating a Power BI report based on a machine learning model. By the end of this tutorial, you'll be able to:

  • Score machine learning models (deployed using Azure Machine Learning) in Power BI.
  • Connect to an Azure Machine Learning model in the Power Query Editor.
  • Create a report with a visualization based on that model.
  • Publish that report to the Power BI service.
  • Set up scheduled refresh for the report.

Prerequisites

Before starting this tutorial, you need to:

Create the data model

Open Power BI Desktop and select Get Data. In the Get Data dialog box, search for web. Select the Web source > Connect.

In the From Web dialog box, copy and paste the following URL in the box:

txt

https://www4.stat.ncsu.edu/~boos/var.select/diabetes.tab.txt

Select OK.


Select Transform data to open the Power Query Editor window.


In the Home ribbon of the Power Query Editor, select the Azure Machine Learning button.

After signing in to your Azure account using single sign-on, you see a list of available services. Select the my-sklearn-service that you created from the Train and deploy a machine learning model tutorial.


Power Query populates the columns automatically for you. You remember that in our schema for the service, we had a Python decorator that specified the inputs. Select OK.

Selecting OK calls the Azure Machine Learning service. It triggers a warning on data privacy for both the data and the endpoint.


Screenshot showing privacy warning.


Select Continue. In the next screen, select Ignore Privacy Levels checks for this file > Save.


Once the data is scored, Power Query creates an additional column named AzureML.my-diabetes-model.

The data that the service returns is a list.

 Note

If you deployed a designer model, you see a record.


To get the predictions, on the Transform ribbon select the Expand column button > Expand to New Rows.


After the expansion, you see the predictions in the AzureML.my-diabetes-model column.

Follow these next steps to finish cleaning up your data model.

  1. Rename the AzureML.my-diabetes-model column to predicted.
  2. Rename the Y column to actual.
  3. Change the type of the actual column: Select the column, then on the Transform ribbon select Data Type > Decimal Number.
  4. Change the type of the predicted column: Select that column, then on the Transform ribbon select Data Type > Decimal Number.
  5. On the Home ribbon, select Close & Apply.

Create a report with visualizations

Now you can create some visualizations to show your data.

  1. In the Visualizations pane, select a Line chart.
  2. With the line chart visual selected:
  3. Drag the AGE field to the Axis.
  4. Drag the actual field to Values.
  5. Drag the predicted field to Values.

Resize the line chart to fill the page. Your report now has a single line chart with two lines, one for the predicted and one for the actual values, distributed by age.

Publish the report

You can add more visualizations if you wish. In the interest of brevity, in this tutorial we'll publish the report.


  1. Save the report.
  2. Select File > Publish > Publish to Power BI.
  3. Sign in to the Power BI service.
  4. Select My Workspace.
  5. When the report is published successfully, select the Open in Power BI link. The report opens the report in Power BI in your browser.

Enable datasets to refresh

In a scenario where the data source is refreshed with new data to score, you need to update your credentials so the data can be scored.


In My Workspace in the Power BI service, in the black header bar, select More options (...) > Settings > Settings.

Select Datasets, expand Data source credentials, then select Edit Credentials.

Follow the instructions for both azureMLFunctions and Web. Make sure that you select a privacy level. You can now set a Scheduled refresh of the data. Select a Refresh frequency and Time zone. You can also select an email address where Power BI can send refresh failure notifications.

Select Apply.

 Note

When the data is refreshed, it also sends the data to your Azure Machine Learning endpoint for scoring.

Clean up resources

 Important

You can use the resources that you created as prerequisites for other Azure Machine Learning tutorials and how-to articles.

1. In the Azure portal, select Resource groups on the far left.


2. From the list, select the resource group that you created.


3. Select Delete resource group.

4. Enter the resource group name. Then select Delete.


5. In My Workspace in the Power BI service, delete the report and the related dataset. You don't need to delete Power BI Desktop or the report on your computer. Power BI Desktop is free.

Next steps

In this tutorial series, you learnt how to set up a schedule in Power BI so that new data can be scored by your scoring endpoint in Azure Machine Learning.