Data Science & ML

Model Drift

/ˈmɒdəl drɪft/

Definition

The degradation of a model's accuracy over time as real-world data patterns diverge from its training distribution.

Example in context

"Our fraud model drifted after new payment methods launched — we retrain monthly to keep accuracy above 95%."

Related terms

Practice this term

Master Model Drift in context by working through exercises in the Data Science & ML module. You'll see the term used in real engineering scenarios with multiple-choice, fill-in-the-blank, and matching drills.