Definition
Model Validation is the process of evaluating whether a model is reliable, sufficiently accurate and appropriate for its intended use, including its performance on relevant data not used for training.
Human Explanation
It checks whether a model has learned relationships that remain useful beyond its training examples rather than merely memorising those examples.
Why it Matters
A model must perform adequately under real conditions, not only during development.
Conceptual Boundary
High training performance does not demonstrate model quality. Validation must reflect the intended population, decision, error costs, operating conditions and relevant risks; a single metric is rarely sufficient.
Practical Perspective
Validation can reveal overfitting, underfitting, poor calibration, unstable behaviour and performance differences across relevant groups. It should continue after deployment because data and operating conditions can change.