precisionrecallkurvor

Precisionrecallkurvor is the term used in some contexts to refer to the precision-recall curve, a graphical tool for evaluating binary classification models. The curve plots precision against recall as the decision threshold is varied, providing insight into how well a model can identify positive instances while limiting false positives.

Precision is defined as the ratio of true positives to all predicted positives (TP / (TP + FP)), while recall (also called sensitivity) is the ratio of true positives to all actual positives (TP / (TP + FN)). By sorting predicted scores and sweeping a threshold from high to low, researchers compute precision and recall at each threshold and plot the resulting points to form the curve.

The precision-recall curve is especially informative when the positive class is rare or imbalanced, where the traditional receiver operating characteristic (ROC) curve may present an overly optimistic view. The area under the curve (AUPRC) or average precision summarizes the curve into a single scalar value, reflecting the overall trade-off between precision and recall across thresholds.

Interpretation and use include comparing models, selecting decision thresholds, and understanding how changes in classification strategy affect the balance between missing positives and misclassifying negatives. Compared with ROC curves, PR curves focus more on the performance of the positive class and are more sensitive to class distribution; a random baseline precision equals the positive class prevalence.

Extensions cover multi-class scenarios (one-vs-rest curves) and alternative constructions of the curve for specific applications. In German-language literature, precisionrecallkurvor serves as the term for this concept, though it is internationally known as the precision-recall curve.