daviesBouldin

function daviesBouldin(ldIn, cl): MetricResult;

Defined in: metrics/labelled.ts:251

Average worst-case overlap between each class and its closest rival.

• Needs: projection only. Labels required.
• Range: [0, ∞), lower is better — the opposite direction to most measures here. Below ~1 means classes are compact relative to their separation.
• Cost: O(N·D + k²) for k classes.

Parameters

Parameter	Type
ldIn	PointsInput
cl	Clusters

Returns

MetricResult

See

Davies & Bouldin, IEEE TPAMI 1 (1979) https://doi.org/10.1109/TPAMI.1979.4766909

Example

import { clusters, daviesBouldin } from "@saehrimnir/sickle";

const cl = clusters(projection, labels);
daviesBouldin(projection, cl).value;  // 0.1603 — unbounded, lower is better