Eval Metrics

Evaluation Metrics Q&A

1Why evaluation metrics matter?
Answer: They quantify model quality and align model choice with business goals.
2Accuracy limitation?
Answer: Misleading on imbalanced datasets.
3Precision definition?
Answer: TP/(TP+FP), correctness of positive predictions.
4Recall definition?
Answer: TP/(TP+FN), ability to find positives.
5F1 score use?
Answer: Harmonic mean of precision and recall when balance is needed.
6ROC-AUC meaning?
Answer: Probability classifier ranks random positive above random negative.
7PR-AUC when useful?
Answer: Better for rare positive class tasks.
8Regression metrics examples?
Answer: MAE, MSE, RMSE, R-squared.
9MAE vs RMSE?
Answer: RMSE penalizes larger errors more heavily.
10Metric selection rule?
Answer: Choose metrics based on class balance and error cost profile.