A library for machine learning that builds predictions using a linear regression.
PHP
Switch branches/tags
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
src Updated description Jan 5, 2018
.gitignore Initial commit Jun 9, 2015
LICENSE Initial commit Jun 8, 2015
README.md Updated description Jan 5, 2018
composer.json Refactoring Jul 2, 2015

README.md

PredictionBuilder

PredictionBuilder is a library for machine learning that builds predictions using a linear regression.

Requirements

This project requires PHP 5.4 or higher because makes use of trait and short array syntax.

Installation

You can install the library by Composer. Add this to your project's composer.json:

"require": {
    "denissimon/prediction-builder": "*"
}

Then run php composer.phar install (or composer install).

Example

use PredictionBuilder\PredictionBuilder;

require_once __DIR__ . '/vendor/autoload.php';

$data = [[1,20],[2,70],[2,45],[3,81],[5,73],[6,80],[7,110]];
$x = 4.5;

// What is the expected y value for a given x value?
try {
    $prediction = new PredictionBuilder($x, $data);
    $result = $prediction->build(); // y = 76.65
} catch (\Exception $e) {
    echo $e->getMessage(), "\n";
}

The returned object has the following properties:

$result->ln_model linear model that fits the data: "29.56362+10.46364x"

$result->cor correlation coefficient: 0.8348

$result->x given x value: 4.5

$result->y predicted y value: 76.65

License

Licensed under the MIT license