Compare validation strategies—Hold-out, K-Fold, Repeated K-Fold, Nested CV, GroupKFold, and Time-Series (walk-forward). See split diagrams, per-fold metrics, pooled ROC/PR, and how leakage can inflate results.
Hold-out vs Cross-Validation: Hold-out is fast for large data; K-fold and Repeated K-fold give more stable estimates on smaller datasets.
Nested CV: Use an inner loop to tune hyperparameters and an outer loop to estimate unbiased performance of the tuned model.
Group & Time-Series: Keep groups intact (GroupKFold) and preserve time order (walk-forward) to avoid leakage.
Leakage: Fit preprocessing inside each training fold. If you flip the leakage toggle, watch metrics become unrealistically optimistic.
Every coffee helps keep the servers running. Every book sale funds the next tool I'm dreaming up. You're not just supporting a site — you're helping me build what developers actually need.
This lab demonstrates proper validation workflows: holdout splits, k‑fold cross‑validation, stratification under imbalance, and leakage checks. It summarizes metrics and variance across folds.