Distribution Drift Detection for ML Monitoring

Intermediate

Master distribution drift detection for production ML monitoring. Learn PSI, KL divergence, Jensen-Shannon divergence, temporal decay weighting, and rolling window analysis to detect when models degrade.

75 min Lab: notebook 5 objectives 4 evidence types
Open Interactive Lab
75
Minutes
5
Objectives
4
Evidence Types
4
Success Criteria

Case Narrative

Learning Objectives

1
Understand why drift detection is critical for ML in production
2
Compute and interpret Population Stability Index (PSI)
3
Apply KL and Jensen-Shannon divergence for distribution comparison
4
Implement temporal decay weighting for streaming analysis
5
Build rolling window drift monitoring systems

Required Evidence

Psi Computation Not collected yet
Divergence Analysis Not collected yet
Temporal Monitoring Not collected yet
Alert Configuration Not collected yet

Case Details

Difficulty
Intermediate
Duration
75 min
Lab Type
notebook
Slug
distribution-drift-detection

Prerequisites

No prerequisites - open to all

Success Criteria

Alert Thresholds Configured Required
Divergences Compared Required
Psi Computed Required
Rolling Monitor Built Required

Tags