README

🚀 A lightweight library for building recommender systems in PHP.

Supports Item-Item Collaborative Filtering via Jaccard Similarity and an extensible architecture to work with different types of storage.

📦 Installation

composer require rustem-kaimolla/recommender-php

(for local development)

git clone https://github.com/rustem-kaimolla/recommender-php.git
cd recommender-php
composer install

✨ Основные возможности

• Item-Item Collaborative Filtering
• Jaccard Similarity (similarity metric)
• InMemory storage for fast tests (extensible for Redis, PostgreSQL and others)
• Clean architecture with interfaces (PSR-4)
• Test coverage (PHPUnit)
• License MIT

⚡ Quick start

<?php

require 'vendor/autoload.php';

use Recommender\Recommender;
use Recommender\Storage\InMemoryStorage;
use Recommender\Similarity\JaccardSimilarity;

$storage = new InMemoryStorage();
$similarity = new JaccardSimilarity();
$recommender = new Recommender($storage, $similarity);

$recommender->addInteraction(1, 10);
$recommender->addInteraction(1, 12);
$recommender->addInteraction(2, 10);
$recommender->addInteraction(2, 11);
$recommender->addInteraction(3, 11);

$recommendations = $recommender->recommendForUser(1, 3);

print_r($recommendations);

📊 Response example

Array
(
    [0] => Array
        (
            [item_id] => 11
            [score] => 0.5
        )
)

User 1 is recommended product 11 because it was purchased by other users who purchased similar products.

🧪 Running tests

composer test

💚 Development plan

• InMemory storage
• Redis Storage
• Popularity-based baseline
• Event weighting assessment (view, like, purchase)
• Similarity Matrix Cache
• Quality metrics (precision@k, recall@k)

📄 License

The project is licensed under MIT License.

🤝 Contribution

If you want to make a pull request or share an idea, you are always welcome! 🚀