A business stakeholder wants a regression metric that is easy to explain as the typical absolute error in the same units as the target and is less sensitive to a few extreme outliers than RMSE. Which metric is the best choice?
Answer & explanation
Correct answer: E
WHY E: MAE is in the original target units, easy to explain as the typical absolute error, and less sensitive to extreme errors than RMSE.
WHY NOT A: R-squared does express the fraction of variance explained, but it is unitless, so it is not read in the target units.
WHY NOT B: Squaring residuals amplifies large misses, so RMSE is more sensitive to outliers, not less.
WHY NOT C: Log loss scores predicted probabilities in probabilistic classification; it is not a metric for continuous targets.
WHY NOT D: F1 is a classification metric built on precision and recall; treating numeric values as class labels does not make it a regression metric.