CERTIFIED-MACHINE-LEARNING-PROFESSIONAL Question #36: Real Exam Question with Answer & Explanation

Question

Which of the following describes concept drift?

Options

• AConcept drift is when there is a change in the distribution of an input variable
• BConcept drift is when there is a change in the distribution of a target variable
• CConcept drift is when there is a change in the relationship between input variables and target
• DConcept drift is when there is a change in the distribution of the predicted target given by the
• ENone of these describe Concept drift

Explanation

Option D correctly captures concept drift as a shift in P(Y|X) - the conditional distribution of the target given the inputs - meaning the model's learned mapping becomes stale even if the input data looks similar (e.g., customer "buy" behavior changes after an economic shock, so the same features now predict differently).

Why the distractors are wrong:

• A describes data drift (or covariate shift): P(X) changes, but the true relationship between X and Y may still hold.
• B describes label drift (prior probability shift): the marginal distribution P(Y) shifts, which is distinct from the conditional relationship changing.
• C sounds close, but it's phrased too loosely - "relationship" can imply correlation or feature importance, whereas concept drift specifically refers to the conditional probability distribution P(Y|X) changing.
• E is eliminated because D is a valid description.

Memory tip: Think of the word "concept" as the rule the model learned (e.g., "these features predict churn"). Concept drift means the rule itself changed - not just who shows up (data drift) or how often the outcome happens (label drift), but how inputs map to outcomes.

No community discussion yet for this question.