38. Overfitting and Underfitting in Machine Learning
38.5 Practice, Tip and Close
Why. Close when you can define over/underfit and name two fixes each.
How.
Chapter practice
- Take any classifier from Chapters 34–35.
- Print train vs test accuracy.
- If the gap is large, reduce complexity and re-check.
- Write four bullets: overfit definition, underfit definition, one fix each.
Thodkyaat sangaycha tar (quick recap)
- Overfit = memorised train; weak on new rows
- Underfit = too weak on train and test
- Compare train vs test metrics
- Fixes: data, features, simpler/richer model, honest split
Samajla ka? Next: Chapter 39 — End-to-end ML project (India fictional dataset).