An analyst has a sufficient volume of data to perform a 3-way partition of the data into training, validation, and test sets to perform honest assessment during the model building process. What is the

Question

Options

• ATo provide a unbiased measure of assessment for the final model.
• BTo compare models, and select and fine-tune the final model.
• CTo reduce total sample size to make computations more efficient.
• DTo build the predictive models.

• A
  86% (12)
• B
  7% (1)
• C
  7% (1)

Explanation

Option A is correct because the test set is held out entirely until after model selection and tuning are complete, meaning it has never influenced any modeling decisions - this makes it a truly unbiased, "honest" measure of how the final model will perform on unseen data.

Why the distractors are wrong:

• B describes the role of the validation set, which is used to compare candidate models and tune hyperparameters.
• C is nonsensical - splitting data into three sets increases bookkeeping complexity and reduces the data available for training, not computational efficiency.
• D describes the role of the training set, which is what the model actually learns from.

Memory tip: Think of the three sets as a job interview process - the training set is the resume (builds the candidate), the validation set is the interview rounds (compares and selects the best candidate), and the test set is the 90-day performance review (unbiased final verdict after the decision is already made).

No community discussion yet for this question.