ROC Curve
A plot showing the True Positive Rate vs False Positive Rate across all classification thresholds.
The ROC curve shows TPR vs FPR across all thresholds – AUC summarizes classification performance in one number.
Explanation
The ROC curve visualizes the tradeoff between sensitivity and specificity. The area under the curve (AUC) summarizes overall performance in one number (0.5 = random, 1.0 = perfect).
Marketing Relevance
ROC curve and AUC are standard for classification model comparisons – threshold-independent and intuitively communicable.
Example
A fraud detector with AUC=0.95 can detect 90% of fraud cases at only 5% false positive rate.
Common Pitfalls
ROC can be overly optimistic with severe class imbalance. PR curve is more informative in such cases.
Origin & History
The ROC curve was developed during WWII for radar signal detection and became an ML standard in the 1990s.
Comparisons & Differences
ROC Curve vs. PR-Kurve
ROC shows TPR vs FPR; PR curve shows precision vs recall. PR is more informative with class imbalance.