Development build. This page describes main, not a released package. The latest published Lodestar.Metrics is 0.4.0 — read its documentation.

Home › Metrics › Classification metrics

Draws the ROC curve — sklearn.metrics.roc_curve.

public static RocCurve Compute(ReadOnlySpan<int> yTrue, ReadOnlySpan<double> yScore, int posLabel = 1, ReadOnlySpan<double> sampleWeight = default, bool dropIntermediate = true)

Parameters — yTrue is the true labels, one per sample. yScore is a score per sample: the higher, the more the model believes posLabel. posLabel is the label counted as positive, 1 by default, where scikit-learn infers it. sampleWeight is one weight per sample, or empty. dropIntermediate drops points the curve does not bend at; true here, matching the reference's default for this curve and not for the other two.

Returns — a RocCurve whose FalsePositiveRate, TruePositiveRate and Thresholds are three parallel arrays of equal length, the first point being the origin at an infinite threshold.

Exceptions — ArgumentException when the inputs disagree in length, are empty, hold a score or weight that is not finite, or weigh every sample zero. A zero-weight sample is dropped before the thresholds form, as scikit-learn drops it.

Example — four samples, and the area under what it draws.

using Lodestar.Metrics;

int[] truth = [0, 0, 1, 1];
double[] scores = [0.1, 0.4, 0.35, 0.8];

RocCurve curve = RocCurve.Compute(truth, scores);
int points = curve.Thresholds.Count;  // => 5

using Lodestar.Metrics;

int[] truth = [0, 0, 1, 1];
double[] scores = [0.1, 0.4, 0.35, 0.8];

RocCurve curve = RocCurve.Compute(truth, scores);
double[] x = [.. curve.FalsePositiveRate];
double[] y = [.. curve.TruePositiveRate];

double area = Auc.Trapezoid(x, y);  // => 0.75

That is RocAuc.Score on the same input, to the last bit — an invariant the test suite asserts over every fixture of the frozen corpus, because no oracle can state it.

Remarks — a class the reference never sees is absent from the denominator, and the rate is NaN rather than a division by zero, which is what scikit-learn warns about and returns.

Applies to — net10.0, netstandard2.0.

See also — PrecisionRecallCurve.Compute, DetCurve.Compute, Auc.Trapezoid, RocAuc.Score, the Python equivalence table.