Model Selection and Cross-Validation

Model selection means choosing between competing workflows, not just algorithms. This lecture covers k-fold cross-validation in detail, hyperparameter tuning with GridSearchCV, choosing the right evaluation metric, and how to report model selection results. We emphasise that a reproducible selection rule is more valuable than an opaque “best model” claim.

Interactive: K-fold cross-validation

5-fold cross-validation: each iteration holds out a different fold for validation while training on the remaining four.