Content-Based Recommendation PHP (based on http://www.codediesel.com/php/item-based-collaborative-filtering-php/)

content_based.php

<?php

/**
 * PHP item based filtering
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * @package   PHP item based filtering
 */
class ContentBasedRecommend extends Recommend
{
	const USER_ID = '__USER__';
	protected $data;

	function __construct($user, $objects)
	{
		$this->data[self::USER_ID] = $this->processUser($user);
		$this->data = array_merge($this->data, $this->processObjects($objects));
	}

	public function getRecommendation()
	{
		$result = [];

		foreach ($this->data as $k => $v) {
			if($k !== self::USER_ID) {
				$result[$k] = $this->similarityDistance($this->data, self::USER_ID, $k);
			}
		}

		arsort($result);
		return $result;
	}

	protected function processUser($user)
	{
		$result = [];

		foreach ($user as $tag) {
			$result[$tag] = 1.0;
		}

		return $result;
	}

	protected function processObjects($objects)
	{
		$result = [];

		foreach ($objects as $object => $tags) {
			foreach ($tags as $tag) {
				$result[$object][$tag] = 1.0	;
			}
		}

		return $result;
	}
}

index.php

<?php

/**
 * PHP item based filtering
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * @package   PHP item based filtering
 */

require_once 'recommend.php';
require_once 'content_based.php';

$objects = [
	'The Matrix' => ['Action', 'Sci-Fi'],
	'Lord of The Rings' => ['Adventure', 'Drama', 'Fantasy'],
	'Batman' => ['Action', 'Drama', 'Crime'],
	'Fight Club' => ['Drama'],
	'Pulp Fiction' => ['Drama', 'Crime'],
	'Django' => ['Drama', 'Western'],
];

$user = ['Drama', 'Crime'];

$engine = new ContentBasedRecommend($user, $objects);

var_dump($engine->getRecommendation());

recommend.php

<?php

/**
 * PHP item based filtering
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * @package   PHP item based filtering
 */

class Recommend {

    
    public function similarityDistance($preferences, $person1, $person2)
    {
        $similar = array();
        $sum = 0;
        $alpha = 0;
        $beta = 0;
    
        foreach($preferences[$person2] as $key=>$value)
        {
            $alpha += $preferences[$person2][$key] * $preferences[$person2][$key];
        }

        foreach($preferences[$person1] as $key=>$value)
        {
            $beta += $preferences[$person1][$key] * $preferences[$person1][$key];
            if(array_key_exists($key, $preferences[$person2])) {
                $sum += $preferences[$person2][$key] * $preferences[$person1][$key];
            }
        }
        
        if($alpha === 0 || $beta === 0)
            return 0;

        return  $sum / sqrt($alpha) / sqrt($beta);     
    }
    
    
    public function matchItems($preferences, $person)
    {
        $score = array();
        
            foreach($preferences as $otherPerson=>$values)
            {
                if($otherPerson !== $person)
                {
                    $sim = $this->similarityDistance($preferences, $person, $otherPerson);
                    
                    if($sim > 0)
                        $score[$otherPerson] = $sim;
                }
            }
        
        array_multisort($score, SORT_DESC);
        return $score;
    
    }
    
    
    public function transformPreferences($preferences)
    {
        $result = array();
        
        foreach($preferences as $otherPerson => $values)
        {
            foreach($values as $key => $value)
            {
                $result[$key][$otherPerson] = $value;
            }
        }
        
        return $result;
    }
    
    
    public function getRecommendations($preferences, $person)
    {
        $total = array();
        $simSums = array();
        $ranks = array();
        $sim = 0;
        
        foreach($preferences as $otherPerson=>$values)
        {
            if($otherPerson != $person)
            {
                $sim = $this->similarityDistance($preferences, $person, $otherPerson);
            }
            
            if($sim > 0)
            {
                foreach($preferences[$otherPerson] as $key=>$value)
                {
                    if(!array_key_exists($key, $preferences[$person]))
                    {
                        if(!array_key_exists($key, $total)) {
                            $total[$key] = 0;
                        }
                        $total[$key] += $preferences[$otherPerson][$key] * $sim;
                        
                        if(!array_key_exists($key, $simSums)) {
                            $simSums[$key] = 0;
                        }
                        $simSums[$key] += $sim;
                    }
                }
                
            }
        }

        foreach($total as $key=>$value)
        {
            $ranks[$key] = $value / $simSums[$key];
        }
        
    array_multisort($ranks, SORT_DESC);    
    return $ranks;
        
    }
   
}

hei, what metode did u use for find similarity, i think that not eucledian in return

what method you use to find similarityDistance?

What calculation was used to calculate the similarity distance in this code?

Never received a notification about this '^^

For anyone in the future, I used Cosine.

Hi, can I know what does it mean by

$result[$object][$tag] = 1.0

1.0 for what?

Thank You.

@nrlsyhmina If I still remember, the idea is that to represent an object X with the tags a, b, and d, in a universe with the tags a, b, c, d, and e, I put one for the tags that my object has and zero for the rest. So X = (1,1,0,1,0). Somewhere else I check if the key exists and, if not, I put a zero.