classificationError

function classificationError(
   ldIn,
   labels,
   k?,
   knn?): MetricResult;

Defined in: metrics/labelled.ts:132

Share of points whose label is not the majority among their visual neighbours.

The decision-rule counterpart of neighborhoodHit: it asks whether a viewer reading the plot would guess right, rather than how mixed the neighbourhood is.

• Needs: projection only. Labels required.
• Range: [0, 1], lower is better.
• Cost: O(N²·D).

k defaults to 20. Ties count as correct — a tie means the picture is genuinely ambiguous there — so the rule reads slightly optimistically.

Parameters

Parameter	Type	Default value
ldIn	PointsInput	undefined
labels	readonly unknown[]	undefined
k	number	20
knn?	Int32Array<ArrayBufferLike>	undefined

Returns

MetricResult

Example

import { classificationError } from "@saehrimnir/sickle";

// k-NN misclassification rate, k = 20 by default. Ties count as correct, so
// this reads slightly optimistically.
classificationError(projection, labels).value;  // 0 — lower is better