Question 34
Domain 3: Model DevelopmentWhy is selecting the correct primary metric still critical even when AutoML handles the search over candidate models?
Correct answer: A
Explanation
AutoML “facilitates model/feature selection,” but it still evaluates and ranks candidates using the metric you choose. If that primary metric is poor, the search can surface the wrong winner because model selection is driven by that criterion.
Why each option is right or wrong
A. Because AutoML ranks models by the chosen metric, so a poor metric can surface the wrong winner
AutoML does not choose a model arbitrarily; it evaluates candidate runs and orders them by the metric you specify, so the selected winner is only as good as that objective. In Databricks AutoML, the primary metric is the criterion used to compare runs during training, so if you pick an unsuitable metric for the business problem, the top-ranked model can be the wrong one even though the search itself was automated.
B. Because AutoML ignores metrics completely
AutoML evaluates candidate models using metrics; it does not ignore metrics.
C. Because metrics matter only for regression and not classification
Metrics are used for both classification and regression model evaluation.
D. Because the metric is used only after deployment
The primary metric guides model selection during training, tuning, and evaluation.