What is concept drift in time series and how does it impact forecasting models?

Concept drift occurs when the statistical properties of a time series change over time, degrading model performance.

Concept drift refers to a shift in the underlying data distribution generating the time series. This breaks assumptions learned during training, especially in stationary or supervised models. Drift can be sudden (structural breaks), gradual (slow evolution), or recurring (seasonal regime shifts). Handling drift requires retraining, online learning, or adaptive models.