Creating real-time predictions

Use your ML deployment to predict future outcomes on new data. You create real-time predictions using the real-time prediction endpoint in the Machine Learning API.

Predictions can be made in real time, such as real-time decisions about customer discounts at checkout. When predictions are generated, you can load the predictive insights into a Qlik Sense app. This lets you visualize and interact with the data and create what-if scenarios.

Creating real-time predictions with the API

The Real-time predictions pane in the ML deployment interface gives you access to the real-time prediction endpoint in the Machine Learning API.

The real-time prediction endpoint s a two-way communication between AutoML and other capabilities in Qlik Cloud, including Qlik Sense and Automations, as well as external applications. You can use the endpoint to programmatically make predictions by passing data to a model and retrieve the prediction results in real time.

Requirements

• An API key is required to use the real-time prediction endpoint. A user must have the Developer role in the tenant to generate an API key.

  For more information about the prediction API, see Machine Learning API.

• The source model for the ML deployment you are using must be activated for making predictions. For more information, see:

  • Approving deployed models

  • Working with model approval as an administrator

• You need the correct permissions for working with ML deployments and predictions. This includes security roles and space roles. See Working with ML predictions.

Viewing data drift and prediction event details

After you run a real-time prediction, open the ML deployment and switch to the Data drift monitoring pane. In this view you can evaluate:

• The level of data drift for each feature involved in the prediction. The comparison is performed between the data you send to the AutoML real-time prediction API and the training dataset.

• Details about the prediction event, such as whether it succeeded or failed, and how many predictions it generated.

For more information, see Monitoring performance and usage of deployed models.