K-Means-TS

💹 K-means and k-means++ clustering implementation. A Typescript rewrite of Skmeans-JS

Quick Start

Installation

npm i kmeans-ts

Importation

import kmeans from "kmeans-ts";

If you want to access the interfaces or utilities within the package, use

import { KMeans, Vectors, Utils } from "kmeans-ts";

Implementation

var input_data: Array<Array<number>> = [
	[1, 12, 14, 4, 25, 35, 22, 3, 14, 5, 51, 2, 23, 24, 15],
	[7, 34, 15, 34, 17, 11, 34, 2, 35, 18, 52, 34, 33, 21],
	[5, 19, 35, 17, 35, 18, 12, 45, 23, 56, 23, 45, 16, 3]
];
var output: Array<Array<number>> = kmeans(input_data, 3, "kmeans");

Returns

{
	"iterations": 1,
	"k": 3,
	"indexes": [2,1,0],
	"centroids": [
		[5,19,35,17,35,18,12,45,23,56,23,45,16,3,0],
		[7,34,15,34,17,11,34,2,35,18,52,34,33,21,0],
		[1,12,14,4,25,35,22,3,14,5,51,2,23,24,15]
	]
}

Functionality & Params

| Param | Description | Sample Type | Required | | ------------ | ------------------------------------------------------------------------------------------------------------------------------ | --------------------------------------- | -------- | | Input Data | Array of values to be clustered. Can be multi-dimensional | Array<number>, Array<Array<number>> | Yes | | K | Num clusters | number | Yes | | Centroids | Initializes centroids. Kmeans for random, Kmeans++ for the K-means++ algorithm. Will attempt to find them if not provided. | String | Optional | | Iterations | Max num of iterations. Default is 10000 | number | Optional |

Returns the following object:

| Return value | Description | Sample type | | ------------ | ----------------------------------------------- | ---------------------- | | Iterations | Num iterations undergone | number | | K | Num clusters | number | | Centroids | Centroid values for each cluster | Array<number> | | Indexes | Index of centroid for each value of input array | Array<Array<number>> |

Further Examples

// K-means w/ 4 clusters & random centroid initialization
var kmeans: KMeans = kmeans(input_data, 4, "kmeans");

// K-means w/ 3 clusters & initial centroids included
var kmeans: KMeans = kmeans(input_data, 3, [
	[3, 1, 5],
	[7, 2, 6],
	[3, 8, 6]
]);

// K-means++ w/ 5 clusters
var kmeans: KMeans = kmeans(input_data, 5, "kmeans++");

// K-means w/ 7 clusters, random centroids, and 15 max iterations
var kmeans: KMeans = kmeans(input_data, 7, null, 15);

K-Means-TS can be seen in MTG-Meta-TS

Development Setup

Simply clone the repository, then if you would like to generate a new ts-config run

--ts-config init

This will create a tsconfig.json file. If you are using VSCode, enter Ctrl-Shift-B and then tsc:watch, which will auto-compile TS to JS. You can also use tsc <filename> to compile from ts to js.

This project uses tsdx for compilation and minification. You can run that with npm start

To test this project, you can navigate to /example and run the testing ground with either ts-node testing_ground.ts, or by compiling it to JS and then running it in the terminal with node testing_ground.js

Alternatively, you can install the awesome VSCode extension Code Runner, which is very convenient

Contributing

1. Fork K-Means-TS here
2. Create a feature branch (git checkout -b feature/fooBar)
3. Commit your changes (git commit -am 'Add some fooBar')
4. Push to the branch (git push origin feature/fooBar)
5. Create a new Pull Request

Meta

Adapted from @Solzimer's Skmeans-JS by @GoldinGuy

Distributed under the MIT license. See LICENSE for more information.