In plain English
Overfitting means a model memorized its examples and fails on new ones. You spot it by comparing the training score with the score on fresh data. Drift means the world changed after the model was built. Keep a frozen test set and run it on a schedule.
- Overfitting
- Memorizing examples instead of learning the pattern.
- Drift
- Inputs or answers changing after the model was built.
- Test set
- Cases the model never trained on.
Picture it
A drift watch routine
1
Keep a frozen test set
The same cases, every time
2
Run it on a schedule
Weekly, or after any change
3