K-Means Clustering Library

This is a simple K-Means clustering library written in JavaScript. It includes an implementation of the K-Means++ initialization algorithm and silhouette scoring to determine the best number of clusters.

Features

• K-Means clustering with K-Means++ initialization
• Silhouette scoring to optimize the number of clusters

Installation

npm install k-means-clustering

Usage

1. K-Means Clustering

Use kMeans to cluster vectors into k groups:

const { kMeans } = require('k-means-clustering');

const vectors = [
  [1, 2],
  [3, 4],
  [5, 6],
  [8, 9],
  [10, 11]
];

const k = 2;
const { centroids, clusters } = kMeans(vectors, k);

console.log('Centroids:', centroids);
console.log('Clusters:', clusters);

2. Find Best K Using Silhouette Score

Use manyKMeansWithSilhouette to test different values of k and pick the one with the highest silhouette score:

const { manyKMeansWithSilhouette } = require('k-means-clustering');

const bestResult = manyKMeansWithSilhouette(vectors, 2, 5);

console.log('Best Centroids:', bestResult.centroids);
console.log('Best Clusters:', bestResult.clusters);

3. Utility Functions

• calculateCentroid(cluster) – Computes the centroid of a cluster.
• initalizeKMeans(vectors, k) – Initializes centroids using K-Means++ method.

Error Handling

Currently error handling has not been implemented.

Please ensure that vectors are of the same dimension (not 0). k should be a valid value for the number of vectors

Example

Sample input and output:

const vectors = [
  [1, 2],
  [3, 4],
  [5, 6],
  [8, 9],
  [10, 11]
];

const k = 2;
const result = kMeans(vectors, k);

console.log(result.centroids); // [[4, 5], [9, 10]]
console.log(result.clusters); // [[[1, 2], [3, 4], [5, 6]], [[8, 9], [10, 11]]]

License

This project is licensed under the MIT License.