A gradient-boosting regressor that predicts delivery times for an online food-delivery platform was trained on six months of historical orders. Two months after deployment, a new municipal traffic law lowers the maximum speed limit from 35 mph to 25 mph on all urban streets. The distributions of the model's input features (order size, time of day, restaurant-to-customer distance, day of week) remain statistically indistinguishable from the training set, yet the model's residuals become consistently positive and the mean absolute error doubles. Which primary cause of model drift best explains this behaviour?
Information about the target variable leaking into the feature set (data leakage)
A covariate shift in the input feature distributions (data drift)
Random measurement noise in the performance metric (irreducible error)
A shift in the relationship between features and target caused by the external policy change (concept drift)
The speed-limit change alters how long drivers actually take to complete deliveries, so the functional relationship between the input features and the target variable (delivery time) has changed even though the feature distributions themselves have not. This is the textbook definition of concept drift. Data drift (covariate shift) would require a measurable change in the input feature distributions, which monitoring rules out. Data leakage would have produced unrealistically good performance during both training and initial deployment, not a sudden post-law degradation. Random measurement noise increases error variance but would not introduce a systematic positive bias in residuals or double the MAE.
Ask Bash
Bash is our AI bot, trained to help you pass your exam. AI Generated Content may display inaccurate information, always double-check anything important.
What is concept drift in machine learning?
Open an interactive chat with Bash
How does concept drift differ from data drift (covariate shift)?
Open an interactive chat with Bash
How can one detect and mitigate concept drift in machine learning models?