Machine Learning
Which graph plots the true positive rate against the false positive rate at various classification thresholds?
ROC curve.
Card 1 of 259. Answer: ROC curve.
ROC curve.
Card 1 of 259. Answer: ROC curve.
About · Quizzes · History of AI · To suggest new questions or report an error send me a dm.
The ROC curve, short for receiver operating characteristic. It plots the true positive rate, TPR = TP / (TP + FN), against the false positive rate, FPR = FP / (FP + TN), as the decision threshold varies. The name comes from signal detection theory, developed for radar receivers around World War II.
Each threshold gives one point. At a very high threshold nothing is flagged, so the curve starts at (0, 0); at a very low one everything is, giving (1, 1). A model might reach TPR 0.80 at FPR 0.10 with a 0.5 threshold, and TPR 0.95 at FPR 0.30 with 0.2. Random guessing follows the diagonal; a perfect model hugs the top-left corner.
The usual follow-up is AUC, the area under the curve: 0.5 for random, 1.0 for perfect, and equal to the probability that a random positive is scored above a random negative. On heavily imbalanced data a small FPR can still mean many false positives, so a precision-recall curve is often more informative.