Question 5
Domain 3: Model DevelopmentWhat is a common advantage of cross-validation over a single train-validation split?
Correct answer: A
Explanation
Cross-validation is preferred because the guide says it can “perform cross-validation as a part of model fitting” and asks to “describe the benefits and downsides of using cross-validation over a train-validation split.” Evaluating multiple folds reduces dependence on one split, so it provides a more stable estimate of model performance.
Why each option is right or wrong
A. It provides a more stable estimate by evaluating multiple folds
Cross-validation trains and scores the model on several different folds, so the performance estimate is averaged across multiple validation sets rather than hinging on one arbitrary split. In the Databricks exam guide, this is the point of the objective to “describe the benefits and downsides of using cross-validation over a train-validation split” and to “perform cross-validation as a part of model fitting,” because repeated fold evaluation reduces variance in the estimate.
B. It always trains fewer models
Cross-validation usually trains more models, especially when combined with grid search and multiple folds.
C. It removes the need for a metric
Model evaluation still requires metrics such as F1, Log Loss, ROC/AUC, RMSE, MAE, or R-squared.
D. It can be used only for regression
Cross-validation applies to both classification and regression tasks.