Cross-validation, and the leaks that silently inflate it

Question

Explain k-fold cross-validation, why it is preferable to a single train/test split, and the leakage mistakes that quietly destroy it on financial data.

Accepted Answer

The procedure. Partition the data into k roughly equal folds. For each fold: train on the other k-1 folds, predict the held-out fold, record the error. Average the k errors to estimate out-of-sample performance. Leave-one-out CV (LOOCV) is the extreme k = n. Why it beats one split. A single hold-out uses only one slice for validation, so the estimate of generalization error is noisy and wastes data. CV lets every observation serve in a validation fold exactly once, giving a lower-variance estimate while still using most of the data to train each model. Bias–variance of the estimate itself tra…

Cross-validation, and the leaks that silently inflate it

Your answer

More Statistics questions