Metrics auc

Auc

The area under a curve you already have — the companion to the three curve types, and the one member here that takes points rather than labels.

RocAuc.Score computes the ROC area without ever exposing the curve, and is what to reach for when the area is all you want. This is for a caller who is holding a curve: it integrates whatever points it is given.

The two agree, and a test says so. The trapezoid over RocCurve's own output equals RocAuc.Score on the same input, which is an invariant no oracle can state — it relates two members of this package rather than one of them to Python — so it is asserted over every fixture of the frozen corpus.

It is the wrong reading of a precision-recall curve

Deliberately so. Over PrecisionRecallCurve's points the trapezoid interpolates between two thresholds as though the curve were straight there, and comes out optimistic: 0.7916… where AveragePrecision.Score sums the steps to 0.8333…. Both are kept, both are documented, and a test asserts they disagree.

Members

Member	What it does
Auc.Trapezoid	The trapezoidal area under the points given.