You use Azure Machine Learning to train a model based on a dataset named dataset1. You define a dataset monitor and create a dataset named dataset2…

The DataDriftDetector.run() method is used to execute an on-demand data drift analysis job. This method allows you to explicitly specify a baseline dataset and a target dataset to compare. In the scenario provided, you would call run() on your defined dataset monitor, passing dataset1 as the baseline and dataset2 as the target to initiate the comparison and calculate the drift magnitude between them.

get: This method retrieves an existing DataDriftDetector object from the Azure Machine Learning workspace by its name; it does not execute a comparison.

backfill: This method is used to run drift detection jobs for a historical period, typically to fill in data for a scheduled monitor, not for a direct comparison of two specific datasets.

update: This method modifies the configuration properties of an existing DataDriftDetector object, such as its schedule or alert settings, but does not trigger a run.

1. Microsoft Azure Official Documentation

DataDriftDetector Class: The official SDK reference clearly documents the run method.

Reference: Microsoft

"azureml.datadrift.DataDriftDetector class

" Azure Machine Learning SDK documentation.

Details: The documentation for the run method states its purpose is to "Run the data drift detector for a given target dataset or time range." It lists parameters such as targetdata and baselinedata

which directly correspond to dataset2 and dataset1 in the question's scenario.

2. Microsoft Azure Official Documentation

"Detect data drift on datasets": This tutorial provides a practical example of using the run method.

Reference: Microsoft

"Tutorial: Detect data drift (preview) on datasets

" Azure Machine Learning documentation.

Details: In the section "Run a data drift analysis

" the sample code demonstrates creating a DataDriftDetector and then calling the run method with a specified starttime

endtime

and computetarget to trigger an on-demand analysis. This confirms run is the correct method for initiating a comparison.