Data Leakage: Scaling Before Splitting, Duplicates Across Folds and Rolling-Origin Checks
This lesson is part of the course
In this lesson
You scale temperatures with minimum and maximum taken from the training rows only, flag records that appear in two folds and check a forecasting rule month by month in time order instead of with shuffled folds.
Buy the course once for $9.99 to unlock every lesson and lab, the final exam and the certificate – lifetime access, no subscription.
Lessons in this course
- Three Piles of Data: Train, Validation & Test Splits from a Shuffle Number Locked
- k-Fold Cross-Validation: Assigning Folds with MOD and Stratifying by Class Locked
- A Toy Model per Fold: k-Nearest-Neighbour Predictions for a Held-Out Day Locked
- Fold Scores: Mean Absolute Error per Fold, the CV Estimate and Its Spread Locked
- Grid Search: A Setting × Fold Table, the Best k and the One-Standard-Error Rule Locked
- Under- and Overfitting: Training vs Validation Error by Model Complexity Locked
- Data Leakage: Scaling Before Splitting, Duplicates Across Folds and Rolling-Origin Checks Locked
- The Final Report: Choose by CV, Retrain on All Pool Days and Score the Test Set Once Locked