Data drift preset

As explained in Machine Learning Platform Engineering

DataDriftPreset is an Evidently preset that generates a report for detecting distributional changes across dataset columns.

DataDriftPreset generates an Evidently report that compares every column in reference and current datasets, helping monitor distributional changes in production ML data. section 6.2.1

The generated report provides:

• Reference and current distributions for each feature. section 6.2.1
• The statistical test used, plus its drift score or p-value. section 6.2.1
• A default Wasserstein-distance threshold of 0.1 for classifying drift. section 6.2.1

Generate and save a report

Define Report(metrics=[DataDriftPreset()]), run it with reference_data and current_data, then save the result with report.save_html("drift_report.html"). section 6.2.1

Figure 6.16 An example of an Evidently data drift report. The panel provides quick information about the number of drifted columns and missing values in the current and reference datasets.

Use the report view to scan drifted columns and missing values across the two datasets. section 6.2.1

How Machine Learning Platform Engineering covers it

How deep it goes

Explains report generation, column comparisons, test scores, and thresholding, but does not detail test configuration.

Examples

Uses a Python report configuration, reference and current data frames, an HTML artifact, and a column-level drift report.

What it assumes

Requires reference and current data snapshots suitable for comparison.

What you can do after

Operate Evidently to generate and export a default data drift report from two datasets.

How it connects

Connects data drift with reference datasets, current datasets, statistical tests, Wasserstein distance, and dashboards.

liveBookmanning.com

Read Machine Learning Platform Engineering on liveBook Get the book at manning.com

Related concepts

• Evidently AIused by · Evidently AI is an open-source Python tool for evaluating, testing, monitoring, and reporting on machine-learning data, models, predictions, and behavior.

This explainer was generated by Manning's knowledge graph from the text of Machine Learning Platform Engineering. The section links point to the parts of the book it draws on.