Concept Drift
Also called Data Drift, Model Drift
A change over time in the relationship between a model's inputs and the correct answers, which makes a once-accurate model gradually less accurate.
What is concept drift?
Concept drift is a shift in the relationship between a model's inputs (features) and the thing it predicts (the label). Because the world changes, a model trained on past data can slowly become less accurate even though nothing about the model itself has changed. The related idea of data drift, where the incoming data changes, is often discussed alongside it.
How does it work?
Teams watch the model's error rate or compare recent data with the data it was trained on, using statistical tests. When a change is flagged, the usual responses are to retrain the model on newer data, or to update it continuously as new data arrives.
What is a simple example?
A fraud detector learned what stolen-card purchases looked like last year. Fraudsters change their methods, so the old patterns stop matching and the model starts missing cases until it is retrained.
Why does it matter?
Accuracy measured at launch does not last. Ongoing model evaluation and periodic retraining are how teams keep deployed models reliable.