Skip to main content
Version: v0.15.0

Create a drift detection test

A drift detection validation test enables you to analyze a model or dataset.

Instructions​

caution

To run a drift detection test on a dataset or model (the datasets of a model), you do not need to define a Connection as Worker. To learn more, see Set a Connection as Worker.

Model​

To create a drift detection test for a model, consider the following instructions:

  1. In the H2O Model Validation navigation menu, click Models.
  2. In the models table, click the name of the model you want to use for the validation test.
  3. In the H2O Model Validation navigation menu, click Drift detection.
  4. Click New test.
  5. In the Test name box, rewrite or leave as-is the validation test name.
  6. Define the drift detection test settings.
    note

    To learn about each setting for a drift detection test, see Settings: Drift detection.

  7. Click Run drift detection.

Dataset​

caution

Before you run a drift detection test on a dataset, ensure the following: If you did not import the dataset to the local platform when you imported the dataset to H2O Model Validation, you need to activate the Connection utilized to import the dataset. In other words, if you unchecked the Import to local platform checkbox, you need to activate the Connection utilized to import the dataset.

To create a drift detection test for a dataset, consider the following instructions:

  1. In the H2O Model Validation navigation menu, click Datasets.
  2. In the datasets table, click the name of the dataset you want to use for the validation test.
  3. In the H2O Model Validation navigation menu, click Drift detection.
  4. Click New test.
  5. In the Test name box, rewrite or leave as-is the validation test name.
  6. Define the drift detection test settings.
    note

    To learn about each setting for a drift detection test, see Settings: Drift detection.

  7. Click Run drift detection.
Feedback