Why Cross-Validation
The train-test split is a lottery; why you need multiple splits to estimate generalisation reliably
Cross-Validation Methods
LOO-CV, LPO-CV, and K-fold; mechanics of the CV loop; when to use each method
K-Fold and Stratified K-Fold
How to choose K; why stratified K-fold is essential on imbalanced data; detailed flow diagram and step-by-step process
Data Leakage and Best Practices
Data leakage trap (scaling before splitting); Pipeline to prevent it; nested CV for hyperparameter tuning; scikit-learn API; comparing models