Why is a held-out test set kept separate during model development?
A.To make training faster
Splitting doesn't speed training — it reserves data.
B.To evaluate the model on data it has never seen — an honest estimate of real-world performance
Unseen-data evaluation = the point of the test set. Correct.
C.To increase the training data volume
The test set is REMOVED from training, not added.
D.To tune hyperparameters directly
Hyperparameter tuning uses the VALIDATION set, not test.
Why this answer
The split: train (learn patterns) → validation (tune/compare models) → test (final unbiased evaluation on unseen data). Testing on training data would only measure memorization — overfitting would go undetected.
Correct answer: B · Last reviewed 2026-02-14
Official AWS references
Practice this interactively
Answer without seeing the solution first in the Machine Learning Principles practice session.
Practice Machine Learning Principles