ROC curve

ROC Curve (Receiver Operating Characteristic)

A plot of true positive rate (recall) against false positive rate at every possible classification threshold — used to visualize and compare binary classifier performance independent of the threshold chosen.

When you'd see it: Binary classification model evaluation — churn, fraud, medical diagnosis. The ROC curve shows how the precision-recall tradeoff evolves as the decision threshold changes. The area under the ROC curve (AUC) summarizes the curve as a single number.

Why it matters: The ROC curve shows a classifier's performance across all thresholds — not just the one you chose to deploy. Comparing ROC curves of multiple models shows which has better underlying discriminative ability. AUC of 1.0 = perfect; 0.5 = random guessing.

Common mistakes: Using ROC/AUC when classes are very imbalanced. With a 1% positive rate, even a poor model can have high AUC by getting negatives right. In highly imbalanced settings, the precision-recall curve is more informative than the ROC curve.

Related AUC, precision, recall, confusion matrix, logistic regression