How to implement collaborative filtering and recommendation systems with PHP

How to implement collaborative filtering and recommendation system with PHP

Collaborative filtering and recommendation system is a very commonly used algorithm and technology, widely used in e-commerce, social media and online services. The collaborative filtering algorithm analyzes a user's behavior and preferences, compares it with the behavior of other users to find similar users, and makes personalized recommendations for users based on these similarities. This article will introduce how to implement collaborative filtering and recommendation systems in PHP.

1. Data preparation
   Before implementing the collaborative filtering and recommendation system, we first need to prepare the data. Generally speaking, we need to have a user-item matrix to record each user's rating or preference for each item. This matrix can be represented by an array. The following is a sample data:

$ratings = [
    'user1' => ['item1' => 4, 'item2' => 3, 'item3' => 5],
    'user2' => ['item1' => 5, 'item2' => 1, 'item3' => 2],
    'user3' => ['item1' => 2, 'item2' => 4, 'item3' => 1],
];

2. Similarity calculation
   In collaborative filtering, we need to calculate the similarity between users. A commonly used similarity calculation method is the Pearson correlation coefficient. The following is a function implemented in PHP to calculate the Pearson correlation coefficient:

function pearson_similarity($ratings, $user1, $user2) {
    $common_items = array_intersect(array_keys($ratings[$user1]), array_keys($ratings[$user2]));
    
    $n = count($common_items);

    $sum1 = $sum2 = $sum1_sq = $sum2_sq = $p_sum = 0;
    
    foreach ($common_items as $item) {
        $rating1 = $ratings[$user1][$item];
        $rating2 = $ratings[$user2][$item];

        $sum1 += $rating1;
        $sum2 += $rating2;

        $sum1_sq += pow($rating1, 2);
        $sum2_sq += pow($rating2, 2);
        
        $p_sum += $rating1 * $rating2;
    }

    $num = $p_sum - ($sum1 * $sum2 / $n);
    $den = sqrt(($sum1_sq - pow($sum1, 2) / $n) * ($sum2_sq - pow($sum2, 2) / $n));

    if ($den == 0) return 0;
    return $num / $den;
}

3. Recommended generation
   With the similarity between users, we can generate based on the similarity recommend. A common approach is the user-based collaborative filtering algorithm. The following is a function of a user-based collaborative filtering recommendation algorithm implemented in PHP:

function user_based_recommendation($ratings, $user, $n = 5) {
    $similarity = array();
    $weighted_sum = array();
    $similarity_sum = array();

    foreach ($ratings as $other_user => $items) {
        if ($other_user == $user) continue;

        $sim = pearson_similarity($ratings, $user, $other_user);
        if ($sim <= 0) continue;

        foreach ($items as $item => $rating) {
            if (!isset($ratings[$user][$item]) || $ratings[$user][$item] == 0) {
                $weighted_sum[$item] += $rating * $sim;
                $similarity_sum[$item] += $sim;
            }
        }
    }

    $rankings = array();
    foreach ($weighted_sum as $item => $weighted_rating) {
        if ($similarity_sum[$item] > 0) {
            $rankings[$item] = $weighted_rating / $similarity_sum[$item];
        }
    }

    arsort($rankings);
    return array_slice($rankings, 0, $n, true);
}

In the above example code, $n represents the number of recommendations to be generated, and the default is 5. user_based_recommendation The function will return an array of items sorted from high to low by recommendation score.

4. Usage Example
   The following is an example of using the above function:

$recommendations = user_based_recommendation($ratings, 'user1', 3);

foreach ($recommendations as $item => $rating) {
    echo "推荐物品：$item, 评分：$rating
";
}

The above example will generate 3 recommended items for user1, and output the result.

Summary:
Through the above steps, we have shown how to use PHP to implement collaborative filtering and recommendation systems. First, we prepared the data of user-item matrix and calculated the similarity between users. Then, personalized recommendations are generated based on the similarity. This is just one implementation method, and actual applications may need to be modified appropriately according to specific needs. I hope this article can help you understand how to use PHP to implement collaborative filtering and recommendation systems.

The above is the detailed content of How to implement collaborative filtering and recommendation systems with PHP. For more information, please follow other related articles on the PHP Chinese website!