
The trade-off between signal efficiency and background rejection as the cut is moved.A receiver operating characteristic (ROC) curve is a graphical plot that illustrates the diagnostic ability of a binary classifier as its discrimination threshold is varied. It plots the true positive rate (sensitivity) against the false positive rate (1 - specificity) at various threshold settings. The curve shows the trade-off between catching true signals (efficiency) and rejecting background noise (rejection). The area under the ROC curve (AUC) is a single scalar measure of overall classifier performance, independent of threshold choice.
Developed during World War II for radar signal detection, later adopted in psychology, medicine, and machine learning.
For each possible threshold value of the classifier's output score, you compute the true positive rate and false positive rate. Plotting these points yields the ROC curve. A perfect classifier has an AUC of 1, while a random classifier has an AUC of 0.5. The curve allows you to choose a threshold that balances sensitivity and specificity according to the application's needs—for example, in particle physics, you might want high background rejection even at the cost of lower signal efficiency.
Used in medical diagnostics to evaluate screening tests, in machine learning to compare classifiers, and in particle physics to optimize event selection cuts.
The ROC curve is invariant to monotonic transformations of the classifier's output score, meaning it only depends on the ranking of scores, not their absolute values.




