DEV Community

Cover image for Machine Learning in PHP: Build a News Classifier Using Rubix ML
Arafat Hossain Ar
Arafat Hossain Ar Subscriber

Posted on

Machine Learning in PHP: Build a News Classifier Using Rubix ML

Introduction

Machine learning is everywhere—recommending movies, tagging images, and now even classifying news articles. Imagine if you could do that within PHP! With Rubix ML, you can bring the power of machine learning to PHP in a way that’s straightforward and accessible. This guide will walk you through building a simple news classifier that sorts articles into categories like “Sports” or “Technology.” By the end, you’ll have a working classifier that can predict categories for new articles based on their content.

This project is perfect for beginners who want to dip their toes into machine learning using PHP, and you can follow along with the complete code on GitHub.

Table of Contents

  1. What is Rubix ML?
  2. Setting Up the Project
  3. Creating the News Classification Class
  4. Training the Model
  5. Predicting New Samples
  6. Final Thoughts

What is Rubix ML?

Rubix ML is a machine learning library for PHP that brings ML tools and algorithms into a PHP-friendly environment. Whether you’re working on classification, regression, clustering, or even natural language processing, Rubix ML has you covered. It allows you to load and preprocess data, train models, and evaluate performance—all in PHP.

Rubix ML supports a wide range of machine learning tasks, such as:

  • Classification: Categorizing data, like labeling emails as spam or not spam.
  • Regression: Predicting continuous values, like housing prices.
  • Clustering: Grouping data without labels, like finding customer segments.
  • Natural Language Processing (NLP): Working with text data, such as tokenizing and transforming it into usable formats for ML.

Let’s dive into how you can use Rubix ML to build a simple news classifier in PHP!

Setting Up the Project

We’ll start by setting up a new PHP project with Rubix ML and configuring autoloading.

Step 1: Initialize the Project Directory

Create a new project directory and navigate into it:

mkdir NewsClassifier
cd NewsClassifier
Enter fullscreen mode Exit fullscreen mode

Step 2: Install Rubix ML with Composer

Make sure you have Composer installed, then add Rubix ML to your project by running:

composer require rubix/ml
Enter fullscreen mode Exit fullscreen mode

Step 3: Configure Autoloading in composer.json

To autoload classes from our project’s src directory, open or create a composer.json file and add the following configuration:

{
    "autoload": {
        "psr-4": {
            "NewsClassifier\\": "src/"
        }
    },
    "require": {
        "rubix/ml": "^2.5"
    }
}
Enter fullscreen mode Exit fullscreen mode

This tells Composer to autoload any classes within the src folder under the NewsClassifier namespace.

Step 4: Run Composer Autoload Dump

After adding the autoload configuration, run the following command to regenerate Composer’s autoloader:

composer dump-autoload
Enter fullscreen mode Exit fullscreen mode

Step 5: Directory Structure

Your project directory should look like this:

NewsClassifier/
├── src/
│   ├── Classification.php
│   └── train.php
├── storage/
├── vendor/
├── composer.json
└── composer.lock
Enter fullscreen mode Exit fullscreen mode
  • src/: Contains your PHP scripts.
  • storage/: Where the trained model will be saved.
  • vendor/: Contains dependencies installed by Composer.

Creating the News Classification Class

In src/, create a file called Classification.php. This file will contain the methods for training the model and predicting news categories.

<?php

namespace NewsClassifier;

use Rubix\ML\Classifiers\KNearestNeighbors;
use Rubix\ML\Datasets\Labeled;
use Rubix\ML\Datasets\Unlabeled;
use Rubix\ML\PersistentModel;
use Rubix\ML\Pipeline;
use Rubix\ML\Tokenizers\Word;
use Rubix\ML\Transformers\TfIdfTransformer;
use Rubix\ML\Transformers\WordCountVectorizer;
use Rubix\ML\Persisters\Filesystem;

class Classification
{
    private $modelPath;

    public function __construct($modelPath)
    {
        $this->modelPath = $modelPath;
    }

    public function train()
    {
        // Sample data and corresponding labels
        $samples = [
            ['The team played an amazing game of soccer'],
            ['The new programming language has been released'],
            ['The match between the two teams was incredible'],
            ['The new tech gadget has been launched'],
        ];

        $labels = [
            'sports',
            'technology',
            'sports',
            'technology',
        ];

        // Create a labeled dataset
        $dataset = new Labeled($samples, $labels);

        // Set up the pipeline with a text transformer and K-Nearest Neighbors classifier
        $estimator = new Pipeline([
            new WordCountVectorizer(10000, 1, 1, new Word()),
            new TfIdfTransformer(),
        ], new KNearestNeighbors(4));

        // Train the model
        $estimator->train($dataset);

        // Save the model
        $this->saveModel($estimator);

        echo "Training completed and model saved.\n";
    }

    private function saveModel($estimator)
    {
        $persister = new Filesystem($this->modelPath);
        $model = new PersistentModel($estimator, $persister);
        $model->save();
    }

    public function predict(array $samples)
    {
        // Load the saved model
        $persister = new Filesystem($this->modelPath);
        $model = PersistentModel::load($persister);

        // Predict categories for new samples
        $dataset = new Unlabeled($samples);
        return $model->predict($dataset);
    }
}
Enter fullscreen mode Exit fullscreen mode

This Classification class contains methods to:

  • Train: Create and train a pipeline-based model.
  • Save the Model: Save the trained model to the specified path.
  • Predict: Load the saved model and predict the category for new samples.

Training the Model

Create a script called train.php in src/ to train the model.

<?php

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

use NewsClassifier\Classification;

// Define the model path
$modelPath = __DIR__ . '/../storage/model.rbx';

// Initialize the Classification object
$classifier = new Classification($modelPath);

// Train the model and save it
$classifier->train();
Enter fullscreen mode Exit fullscreen mode

Run this script to train the model:

php src/train.php
Enter fullscreen mode Exit fullscreen mode

If successful, you’ll see:

Training completed and model saved.
Enter fullscreen mode Exit fullscreen mode

Predicting New Samples

Create another script, predict.php, in src/ to classify new articles based on the trained model.

<?php

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

use NewsClassifier\Classification;

// Define the path to the saved model
$modelPath = __DIR__ . '/../storage/model.rbx';

// Initialize the Classification object
$classifier = new Classification($modelPath);

// Define new samples for classification
$samples = [
    ['The team played an amazing game of soccer, showing excellent teamwork and strategy.'],
    ['The latest programming language release introduces features that enhance coding efficiency.'],
    ['An incredible match between two top teams ended in a thrilling draw last night.'],
    ['This new tech gadget includes features never before seen, setting a new standard in the industry.'],
];

// Predict categories
$predictions = $classifier->predict($samples);

// Display predictions
foreach ($predictions as $index => $prediction) {
    echo "Sample: " . $samples[$index][0] . "\n";
    echo "Prediction: " . $prediction . "\n\n";
}
Enter fullscreen mode Exit fullscreen mode

Run the prediction script to classify the samples:

php src/predict.php
Enter fullscreen mode Exit fullscreen mode

The output should show each sample text with its predicted category.

Final Thoughts

With this guide, you’ve successfully built a simple news classifier in PHP using Rubix ML! This demonstrates how PHP can be more versatile than you might think, bringing in machine learning capabilities for tasks like text classification, recommendation systems, and more. The full code for this project is available on GitHub.

Experiment with different algorithms or data to expand the classifier. Who knew PHP could do machine learning? Now you do.
Happy coding!

Top comments (0)