CERTIFIED-MACHINE-LEARNING-PROFESSIONAL Question #58: Real Exam Question with Answer & Explanation

Question

Which of the following describes label drift?

Options

• ALabel drift is when there is a change in the distribution of the predicted target given by the model
• BNone of these describe label drift
• CLabel drift is when there is a change in the distribution of an input variable
• DLabel drift is when there is a change in the relationship between input variables and target
• ELabel drift is when there is a change in the distribution of a target variable

Explanation

There's an error in this question - the marked answer (C) is actually incorrect by standard ML definitions, and following it would mislead you on an exam.

Option E is the correct definition of label drift: a change in the distribution of the target/label variable (e.g., fraud cases go from 5% to 15% of transactions over time). The "label" in label drift refers to the output label, not an input.

Why each option is wrong (or misidentified):

• C (marked as correct, but wrong): describes feature drift (also called covariate drift or data drift) - a change in input variable distribution, not the label.
• A: describes a change in the model's predictions, which is a downstream consequence of drift, not a definition of label drift.
• D: describes concept drift - a change in the relationship P(Y|X) between inputs and target.
• B: "None of these" is wrong because E is a valid definition.

Memory tip: Think of it this way - the "label" is what you're predicting (the Y). So label drift = Y's distribution shifts, feature drift = X's distribution shifts, and concept drift = the X→Y relationship shifts.

Bottom line: If your exam uses this question, be aware the answer key appears to have swapped label drift (E) with feature/covariate drift (C). Option E is the academically standard definition of label drift.

No community discussion yet for this question.