What is the purpose of AUROC, and when is it particularly useful?

• A. AUROC measures the area under the ROC curve; useful for ranking across thresholds and in imbalanced data.
• B. AUROC is the accuracy at a fixed threshold.
• C. AUROC measures F1 score.
• D. AUROC is the log loss.

Answer: (A)

AUROC measures how well a model can discriminate between classes across all possible decision thresholds. By sweeping the threshold, you plot the true positive rate against the false positive rate and take the area under that curve. This makes AUROC particularly useful when you want to rank predictions or compare models without committing to a single cutoff, and it helps in imbalanced data situations because the measure weighs relative ranking rather than sheer accuracy at one threshold. An AUROC near 1 indicates strong separation between positives and negatives, while around 0.5 means the model is no better than random. In contrast, accuracy at a fixed threshold depends on that specific cutoff; F1 score couples precision and recall at a chosen threshold; and log loss assesses the calibration of predicted probabilities, not ranking across thresholds.