Cross-validation, AIC, and BIC metrics for evaluating predictive performance and model complexity. Balances bias and variance to select the most generalizable model for a given dataset.