Mehdi Imani

Why ROC-AUC Alone Is Insufficient for Highly Imbalanced Data: In-Depth Evaluation of MCC, F2-Score, H-Measure, and AUC-Based Metrics for Rare-Event Binary Classification

2025 · Preprints.org

This study re-evaluates ROC-AUC for binary classification under severe class imbalance (<3% positives). Despite widespread use, ROC-AUC can mask operationally salient differences among classifiers when false-positive and false-negative costs are asymmetric. Using three benchmarks, credit-card …