feraiur/recommender_system_php

Collaborative Recommender Filtering and Apriori Rules Associations for PHP language

dev-master 2020-08-13 16:10 UTC

MIT

• Fernando Sanchez

README

Recommender for PHP is a library that allows to create recommendations based on:

• Collarative Filtering.
• Apriori association rules.

How To

Include the library using composer and then use the example below:

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

use feraiur\recommender_system_php\Predictor;
use feraiur\recommender_system_php\AssociationRules;


// Collaborative Filering

$predictor = new Predictor();
$dataseRatings = [
                [1, 1, 3],
                [2, 2, 5],
                [3, 3, 5],
                [4, 4, 5],
                [5, 5, 5],
                [2, 1, 2],
                [3, 1, 2],
                [4, 1, 2],
                [5, 1, 2],
              ];

try {
    $predictor->setDataset($dataseRatings);
    $recommendations = $predictor->predict();
} catch(Exception $e) {
    var_dump($e->getMessage());
    $recommendations = [];
}


echo("<h3>Recommendations:</h3>");
echo("<p>Some recommendations for users about items.</p>");
var_dump($recommendations);


// Association Rules

$associationRules = new AssociationRules(20, 50);
$datasetTransactions = [
                [1, 3],
                [3, 1],
                [2, 3],
                [3, 4],
                ["i1", 4],
                ["i2", 4],
                [1, 4],
                ["i2", 3],
                ["i2", 3, 4],
                ["i2", 3, "i1"],
              ];

try {
    $associationRules->setDataset($datasetTransactions);
    $supportsK1 = $associationRules->removeUnwantedSupports($associationRules->getSupports());
    $supportsK2 = $associationRules->removeUnwantedSupports($associationRules->getK2Supports($associationRules->getItemsetsWithK2($supportsK1)));
    $finalPairsItemsets = $associationRules->getFinalPairsItemsets($supportsK2);
    $confidenceRules = $associationRules->getConfidenceRules($finalPairsItemsets, $supportsK2, $supportsK1);

} catch(Exception $e) {
    var_dump($e->getMessage());
    $rules = [];
}

echo("<h3>Associations Rules:</h3>");
echo("<p>Ex: People who has item {A} has item {B} too.</p>");
var_dump($confidenceRules);

Datasets format

• Dataset for rating predictions should be an array with 3 integer columns: user_id, item_id and rating value. (Items shuld not be duplicated for the same user).
• Dataset for apriori rules should be an array that contain another array of item ids. (Items should not be duplicated inside the array of each transaction)

Author

Eng. Fernando Sánchez

Unit Test Battery

The unit tests battery is inside the folder "/tests"

• Run php phpunit.phar --verbose RecommendationTest.php