Metrics completeness score

Completeness.Score

Whether every sample of one class landed in the same cluster.

public static double Score(ReadOnlySpan<int> labelsTrue, ReadOnlySpan<int> labelsPred)

Parameters — labelsTrue is the reference partition and labelsPred the one being scored, one label per sample and the same length. The label values carry no meaning: only which samples share one does.

Returns — double in [0, 1]. 1 when no class is split across clusters, 0 when the clustering says nothing about the classes.

Exceptions — ArgumentException when the two labellings disagree in length. An empty input is not an error: it scores 1.

Example — merging two classes into one cluster keeps each class together, and costs nothing here.

using Lodestar.Metrics;

int[] truth = [0, 0, 1, 1];
int[] merged = [0, 0, 0, 0];

double whole = Completeness.Score(truth, merged);   // => 1

Remarks — the mirror of Homogeneity.Score, and literally so: this is that score with the two labellings exchanged. The example scores 1 here and 0 there, which is why neither number means anything on its own.

Applies to — net10.0, netstandard2.0.

See also — Homogeneity.Score, VMeasure.Score, the Python equivalence table.