The method used to split data and evaluate model performance, such as cross-validation or hold-out test sets.