PHP K-Means Library

A PHP implementation of the K-Means clustering algorithm.

Requirements

• PHP 8.0 or higher
• Composer

Installation

Install the library using Composer:

composer require medansoftware/kmeans-algorithm-php

Usage

Include Composer's autoloader:

require 'vendor/autoload.php';

Example 1: Two-dimensional data

This example uses two attributes (A and B) and divides the data into two clusters.

<?php

require 'vendor/autoload.php';

$kmeans = new \Algorithm\KMeans;

$kmeans->setAttributes(array(
    'A',
    'B'
));

$kmeans->setDataFromArgs(1, 1);
$kmeans->setDataFromArgs(2, 1);
$kmeans->setDataFromArgs(4, 3);
$kmeans->setDataFromArgs(5, 4);

$kmeans->setClusterCount(2);

/*
 * Use data points at indexes 0 and 1
 * as the initial centroids.
 */
$kmeans->setCentroid(0, 1);

/*
 * Run K-Means with a maximum of 100 iterations.
 */
$kmeans->setIteration(100);
$kmeans->run();

echo '<h2>Initial Centroids</h2>';
echo '<pre>';
print_r($kmeans->getInitialCentroid());
echo '</pre>';

echo '<h2>Final Centroids</h2>';
echo '<pre>';
print_r($kmeans->getCentroid());
echo '</pre>';

echo '<h2>Iterations</h2>';
echo 'Iterations executed: ' . $kmeans->countIterations();

echo '<h2>Results</h2>';
echo '<pre>';
print_r($kmeans->getAllResults());
echo '</pre>';

Example 2: One-dimensional data

This example uses a single attribute and divides the data into three clusters.

<?php

require 'vendor/autoload.php';

$kmeans = new \Algorithm\KMeans;

$kmeans->setAttributes(array(
    'x'
));

$kmeans->setDataFromArgs(1);
$kmeans->setDataFromArgs(2);
$kmeans->setDataFromArgs(6);
$kmeans->setDataFromArgs(7);
$kmeans->setDataFromArgs(8);
$kmeans->setDataFromArgs(10);
$kmeans->setDataFromArgs(15);
$kmeans->setDataFromArgs(17);
$kmeans->setDataFromArgs(20);

$kmeans->setClusterCount(3);

/*
 * Use data points at indexes 1, 5, and 7
 * as the initial centroids.
 *
 * The selected values are 2, 10, and 17.
 */
$kmeans->setCentroid(1, 5, 7);

/*
 * Run K-Means with a maximum of 100 iterations.
 */
$kmeans->setIteration(100);
$kmeans->run();

echo '<h2>Initial Centroids</h2>';
echo '<pre>';
print_r($kmeans->getInitialCentroid());
echo '</pre>';

echo '<h2>Final Centroids</h2>';
echo '<pre>';
print_r($kmeans->getCentroid());
echo '</pre>';

echo '<h2>Iterations</h2>';
echo 'Iterations executed: ' . $kmeans->countIterations();

echo '<h2>Logs</h2>';
echo '<pre>';
print_r($kmeans->catchLogs());
echo '</pre>';

Initial Centroids

Initial centroids can be selected from existing data points by using their indexes:

$kmeans->setCentroid(0, 1);

An array of indexes can also be used:

$kmeans->setCentroid(array(0, 1));

Alternatively, centroid values can be provided directly:

$kmeans->setCentroid(array(
    array('A' => 1, 'B' => 1),
    array('A' => 5, 'B' => 4)
));

If no centroids are provided, the library automatically selects distinct data points as initial centroids:

$kmeans->setCentroid();

Iterations and Convergence

The maximum number of iterations can be configured with:

$kmeans->setIteration(100);

The algorithm automatically stops when the centroids converge or when the maximum number of iterations is reached.

The number of iterations executed can be obtained using:

$kmeans->countIterations();

You can check whether the algorithm has converged using:

$kmeans->isDone();

Results

Get the final centroids:

$kmeans->getCentroid();

Get the initial centroids:

$kmeans->getInitialCentroid();

Get the cluster logs:

$kmeans->getClusters();

Get all logs:

$kmeans->catchLogs();

Get all results:

$kmeans->getAllResults();

References

• Code
• Article
• Video (Indonesian)

Made with ❤️ + ☕ ~ Agung Dirgantara