Question 39
Domain 4: ML Solution Monitoring, Maintenance, and SecurityA machine learning team sees that the distribution of incoming feature values has remained stable, but the relationship between those features and the target outcome has changed, causing prediction accuracy to decline. Which type of drift has MOST likely occurred?
Correct answer: B
Explanation
Data drift changes the distribution of input data, while concept drift changes the relationship between inputs and the target. When feature distributions stay similar but model accuracy drops because the mapping changes, concept drift is the better diagnosis. — Concept drift vs data drift detection patterns
Why each option is right or wrong
A. Data drift affecting the feature distribution seen by the model
Data drift involves changes in the input data distribution, not a stable distribution with a changed target relationship.
B. Concept drift changing how features relate to the target outcome
The scenario states that incoming feature values remain stable, but prediction accuracy declines because the relationship between features and the target has changed. Under the distinction between concept drift and data drift, that pattern matches concept drift rather than a shift in the feature distribution itself.
C. Data drift caused by lower model accuracy on recent predictions
Lower accuracy alone does not define data drift; the key condition is a change in input distribution.
D. Concept drift caused by a change in the volume of incoming records
Concept drift concerns the input-target relationship, not the number of records received.