stojg\recommend\strategy\Paerson

Complexity

Total Complexity

5

Size/Duplication

Total Lines	50
Duplicated Lines	0 %

Coupling/Cohesion

Components	0
Dependencies	0

Importance

Changes	3
Bugs	1	Features	0

1 Method

Rating	Name	Duplication	Size	Complexity
B	run()	0	39	5

<?php

namespace stojg\recommend\strategy;

/**
 * Use if the data is subject to grade-inflation (different users may be using different scales).
 */
class Paerson
{
    /**
     * Single pass version of the paerson.
     * 
     * @param array $ratings1
     * @param array $ratings2
     *
     * @return float
     */
    public function run($ratings1, $ratings2)
    {
        $numCoRatedItems = 0;
        $dotProduct = 0;
        $rating1Sum = 0;
        $rating1SumSqr = 0;
        $rating2Sum = 0;
        $rating2SumSqr = 0;
        foreach ($ratings1 as $item => $rating1) {
            if (!isset($ratings2[$item])) {
                continue;
            }
            $numCoRatedItems += 1;
            $dotProduct += $rating1 * $ratings2[$item];
            $rating1Sum += $rating1;
            $rating1SumSqr += pow($rating1, 2);
            $rating2Sum += $ratings2[$item];
            $rating2SumSqr += pow($ratings2[$item], 2);
        }

        // There is no correlation at all
        if ($numCoRatedItems == 0) {
            return false;
        }

        $denominator = sqrt(
            ($rating1SumSqr - (pow($rating1Sum, 2) / $numCoRatedItems)) *
            ($rating2SumSqr - (pow($rating2Sum, 2) / $numCoRatedItems))
        );

        if ($denominator == 0) {
            return false;
        }

        // the closer abs(paerson) is to 1 to better correlation is it
        $paerson = ($dotProduct - ($rating1Sum * $rating2Sum / $numCoRatedItems)) / $denominator;

        return 1 - abs($paerson);
    }
}