When can AUROC be misleading?

Question: In highly imbalanced classification, why can AUROC sometimes be misleading? A) Because AUROC only works for regression B) Because AUROC requires balanced mini-batches C) Because AUROC cannot compare rankings D) Because it may look high even when positive-class precision is poor Correct: D Explanation: AUROC measures ranking quality across thresholds, but in highly imbalanced problems, a model can rank reasonably well while still producing many false positives. Precision-recall metrics are often more informative when the positive class is rare. Topic: advanced ML / imbalanced classification / metrics