Reference · Glossary

Cross-validation

Splitting data into **multiple train/test folds** so you estimate performance more reliably than a single lucky split.

When to use

Small tabular datasets, hyperparameter tuning, and student labs before trusting one accuracy number.

When not to

Massive streaming data at scale — simpler holdout or time-based splits may suffice.

Example

5-fold CV: train on 4 chunks, test on the 5th, rotate — report average accuracy across folds.