Skip to content

Click in-app to access the full platform documentation for your version of DataRobot.

Batch Prediction module in pipelines

Availability information

The Batch Prediction module is off by default. Contact your DataRobot representative or administrator for information on enabling the feature.

Feature flag: Enable Pipeline Workspaces in the AI Catalog and Enable Workspace Experimental Modules

The Batch Prediction module, now available for public preview, is the first in a series of modules to bring machine learning capabilites into DataRobot Pipelines. This module allows you to generate batch predictions for an input dataset, as well as perform pre- and post- processing steps in a pipeline.

To add a Batch Prediction module to a pipeline:

  1. Navigate to AI Catalog > Workspaces. You can either click Add new workspace to create a new pipeline, or select an existing pipeline.

  2. To add a Batch Prediction module, hover over the output port of an existing module or click Add new module. The Batch Prediction module must be preceded by another module to generate predictions.

  3. Under Machine Learning on the Add new module screen, select Batch Prediction.

  4. Click the Details tab and select a deployment you want to use to generate predictions from the dropdown.

  5. Fill out the remaining options, described in the table below:

    Option Description
    Include input features Toggle on to include input features when making predictions.
    • To include all input features, select All features.
    • To only include specific input features, select Specific features and begin typing the feature name in the field and select features from the list that appears below.
    Include prediction explanations Toggle on to generate prediction explanations for batch predictions. When enabled, you can define the number of prediction explanations generated, as well as the high and low prediction threshold.
    Track data drift and accuracy for predictions Toggle on to track data drift and accuracy for predictions generated in the pipeline.
  6. You can then use the Batch Prediction module as the input port for another module.

When you run the pipeline, prediction results appear in the Results tab at the bottom of the page.

Updated April 19, 2022
Back to top