Arize AI
Search…
6. Troubleshoot Drift
Drift Tab offers a deep dive into Prediction Drift, Feature Drift, and Data Distribution Change.
Previously, we walked you through setting up real-time drift monitors.
A monitor is triggered and you are alerted that your model has drifted beyond your threshold value. Learn how to troubleshoot what's causing your model to drift from prediction drift to feature drift and data distribution changes.

Navigate to the Drift Tab

Begin troubleshooting drift by accessing the Drift Tab. Access the drift tab via the homepage or within each drift monitor.
Navigate to the Drift Tab - Homepage
From the 'Space Overview' page, click on your model of interest, which will take you to the 'Model Overview' page. From there, click the 'Drift' tab located on the top navigation bar.
Navigate to the 'Drift Tab' from the 'Space Overview' page
Navigate to the Drift Page - Monitor
If you're looking at individual monitors in the Monitors Tab, access the Drift Tab from the 'Troubleshoot Drift' button located on the top right of your drift monitor.
Navigate to the Drift Page from a monitor

Prediction Drift

Prediction Drift refers to a change in the prediction distribution generated by your model. Arize can surface a change in distributions for all model types and problems such as regression, binary, multi-classification, and score models.
Once you're in the Drift Tab, you'll see a Prediction Drift Overtime graph overlaid with traffic and a secondary metric of your choosing. Find and edit details of the environment, version, and time range of your data at the top of the page. Since drift is measured using a reference data set, you'll also see details of your baseline dataset that's used as a comparison.
Begin troubleshooting by clicking on parts of your model that deviate from your threshold on the Prediction Drift curve. From there, uncover changes in your Distribution Comparison and Feature Drift in the cards below.
Distribution Comparisons for values above and below the model baseline

Feature Drift

Once you've identified problem areas in your model to investigate, narrow down on problematic areas to troubleshoot with the Feature Drift card under the Performance Drift Chart. Feature Drift refers to changes in your feature distribution. Since feature drift is very common in the real world, it can happen at any time while your model is in production, yet remain undetected.
The Feature Drift card highlights the most impactful features degrading your model and uncovers granular drift information per feature.
Click into each feature name to look at the feature drift over time chart and the feature's distribution comparison to easily identify missing values, problematic slices, and areas to improve.
Prediction Drift Impact default sorts the features with the most impact on your model
  • Drift (PSI) is a measurement of how much your distribution has drifted.
  • Feature Importance helps you explain why even a small Drift (PSI) can have a significant Drift Impact.

Troubleshooting Tutorial

For a quick-start intro on how to use the drift tab, check out the Troubleshooting Drift Tutorial!

Additional Resources

Questions? Email us at [email protected] or Slack us in the #arize-support channel
Copy link
Outline
Navigate to the Drift Tab
Prediction Drift
Feature Drift
Troubleshooting Tutorial
Additional Resources