kmeans 1.0.0 icon indicating copy to clipboard operation
kmeans: ^1.0.0 copied to clipboard

A Dart library for k-means clustering

package:kmeans #

An implementation of k-means clustering for Dart.

k-means tends to be inflexible, preferring to produce equal-sized clusters. See the wikipedia article for full details.

In this implementation, data points are translated and scaled in each dimension to try to give each parameter equal weight.

By default, clusters are initialized using the k-means++ method. However, repeated trials are still recommended to avoid local minima. The user can also supply an initial set of means.

To evaluate the goodness of a clustering, and to find good values for k when one is not known in advance, this library includes a calculation of the silhouette of a clustering, and uses it in the bestFit() method. The

Usage #

A simple usage example:

import 'package:kmeans/kmeans.dart';

main() {
  var kmeans = KMeans([
    [0.0, 0.0], [1.1, 1.1], [-5.0, -0.50], ...
  var k = 3;
  var clusters = kmeans.compute(k);
  var silhouette = clusters.silhouette;
  print('The clusters have silhouette $silhouette');
  for (int i = 0; i < kmeans.points.length; i++) {
    var point = kmeans.points[i];
    var cluster = kmeans.clusters[i];
    var mean = kmeans.means[cluster];
    print('$point is in cluster $cluster with mean $mean.');
  var bestCluster = kmeans.computeBest(
    minK: 3,
    maxK: 10,

Features and bugs #

Please file feature requests and bugs at the issue tracker.

pub points


unverified uploader

A Dart library for k-means clustering



API reference


BSD-3-Clause (LICENSE)


Packages that depend on kmeans