The dye-house slices were all drawn from winter lots. Spring lots are more pastel. What must be true of the training, validation, and test slices?
Select an answer to reveal the explanation.
Short Explanation
Winter-only slices cannot stand in for pastel spring lots. Every slice used to train, evaluate, tune, and test must represent the data the model will see operationally. Validation and test may not stay winter-only.
Full Explanation
All three slices must represent future operational data. Restricting that duty to training, moving it only to Chapter 5, or ignoring a seasonal shift leaves a non-representative development cycle.