https://github.com/philmod/node-kmeans

Node.js asynchronous implementation of the clustering algorithm k-means.
https://github.com/philmod/node-kmeans

Last synced: 4 days ago
JSON representation

Node.js asynchronous implementation of the clustering algorithm k-means.

• Host: GitHub
• URL: https://github.com/philmod/node-kmeans
• Owner: Philmod
• Created: 2012-10-21T15:43:12.000Z (about 12 years ago)
• Default Branch: master
• Last Pushed: 2022-12-30T17:54:49.000Z (about 2 years ago)
• Last Synced: 2024-12-31T09:13:44.029Z (11 days ago)
• Language: JavaScript
• Homepage:
• Size: 310 KB
• Stars: 102
• Watchers: 6
• Forks: 26
• Open Issues: 14
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# node-kmeans

  Node.js asynchronous implementation of the clustering algorithm k-means

![k-means](http://www.aishack.in/static/img/tut/kmeans-example.jpg)

## Installation

      $ npm install node-kmeans

## Example

```js

// Data source: LinkedIn

const data = [

  {'company': 'Microsoft' , 'size': 91259, 'revenue': 60420},

  {'company': 'IBM' , 'size': 400000, 'revenue': 98787},

  {'company': 'Skype' , 'size': 700, 'revenue': 716},

  {'company': 'SAP' , 'size': 48000, 'revenue': 11567},

  {'company': 'Yahoo!' , 'size': 14000 , 'revenue': 6426 },

  {'company': 'eBay' , 'size': 15000, 'revenue': 8700},

];

// Create the data 2D-array (vectors) describing the data

let vectors = new Array();

for (let i = 0 ; i < data.length ; i++) {

  vectors[i] = [ data[i]['size'] , data[i]['revenue']];

}

const kmeans = require('node-kmeans');

kmeans.clusterize(vectors, {k: 4}, (err,res) => {

  if (err) console.error(err);

  else console.log('%o',res);

});

```

## Inputs

 - **vectors** is a nXm array (n [lines] : number of points, m [columns] : number of dimensions)

 - **options** object:

    - **k** : number of clusters

    - **distance** (optional) : custom distance function returning the distance between two points `(a,b) => number`, *default* Euclidian Distance

    - **seed** (optional) : value that can be provided to get repeatable cluster generation

 - **callback** node-style callback taking error and result argument

## Outputs

An array of objects (one for each cluster) with the following properties:

 - centroid : array of X elements (X = number of dimensions)

 - cluster : array of X elements containing the vectors of the input data

 - clusterInd : array of X integers which are the indexes of the input data

## To do

 - Technique to avoid local optima (mutation, ...)

## Author

Philmod <[email protected]>