A precision-recall curve (or PR Curve) is a plot of the precision (y-axis) and the recall (x-axis) for different probability thresholds.
Proposed by @Al_Grigor
Card 1 of 259. Answer: A precision-recall curve (or PR Curve) is a plot of the precision (y-axis) and the recall (x-axis) for different probability thresholds.
About · Quizzes · History of AI · To suggest new questions or report an error send me a dm.
A precision-recall (PR) curve plots precision on the y-axis against recall on the x-axis as you sweep the classifier's decision threshold from high to low. Each point is one threshold; together they show the tradeoff between catching more positives and raising more false alarms.
The area under it, usually reported as average precision (AP), summarizes the curve in one number. A random classifier's PR curve is a flat line at the positive-class rate, so on a dataset with 1% positives the baseline is 0.01, not 0.5.
PR curves are preferred over ROC curves when positives are rare, as in fraud, disease screening or search. ROC uses the false positive rate, whose denominator is the huge pool of negatives, so it can look excellent while precision is terrible. The PR curve shows that problem directly. It is also how you pick an operating threshold: choose the point whose precision and recall fit the cost of each kind of error.
