Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
0.00% |
0 / 29 |
|
0.00% |
0 / 4 |
CRAP | |
0.00% |
0 / 1 |
| BayesianPersonalizedRanking | |
0.00% |
0 / 29 |
|
0.00% |
0 / 4 |
110 | |
0.00% |
0 / 1 |
| __construct | |
0.00% |
0 / 3 |
|
0.00% |
0 / 1 |
2 | |||
| generateRandomFactors | |
0.00% |
0 / 4 |
|
0.00% |
0 / 1 |
6 | |||
| predict | |
0.00% |
0 / 6 |
|
0.00% |
0 / 1 |
6 | |||
| updateFactors | |
0.00% |
0 / 16 |
|
0.00% |
0 / 1 |
30 | |||
| 1 | <?php |
| 2 | /** |
| 3 | * Jingga |
| 4 | * |
| 5 | * PHP Version 8.1 |
| 6 | * |
| 7 | * @package phpOMS\Business\Recommendation |
| 8 | * @copyright Dennis Eichhorn |
| 9 | * @license OMS License 2.0 |
| 10 | * @version 1.0.0 |
| 11 | * @link https://jingga.app |
| 12 | */ |
| 13 | declare(strict_types=1); |
| 14 | |
| 15 | namespace phpOMS\Business\Recommendation; |
| 16 | |
| 17 | /** |
| 18 | * Bayesian Personalized Ranking (BPR) |
| 19 | * |
| 20 | * @package phpOMS\Business\Recommendation |
| 21 | * @license OMS License 2.0 |
| 22 | * @link https://jingga.app |
| 23 | * @see https://arxiv.org/ftp/arxiv/papers/1205/1205.2618.pdf |
| 24 | * @since 1.0.0 |
| 25 | * |
| 26 | * @todo Implement, current implementation probably wrong |
| 27 | */ |
| 28 | final class BayesianPersonalizedRanking |
| 29 | { |
| 30 | private int $numFactors; |
| 31 | |
| 32 | private float $learningRate; |
| 33 | |
| 34 | private float $regularization; |
| 35 | |
| 36 | private array $userFactors = []; |
| 37 | |
| 38 | private array $itemFactors = []; |
| 39 | |
| 40 | // num_factors determines the dimensionality of the latent factor space. |
| 41 | // learning_rate controls the step size for updating the latent factors during optimization. |
| 42 | // regularization prevents overfitting by adding a penalty for large parameter values. |
| 43 | public function __construct(int $numFactors, float $learningRate, float $regularization) |
| 44 | { |
| 45 | $this->numFactors = $numFactors; |
| 46 | $this->learningRate = $learningRate; |
| 47 | $this->regularization = $regularization; |
| 48 | } |
| 49 | |
| 50 | private function generateRandomFactors() |
| 51 | { |
| 52 | $factors = []; |
| 53 | for ($i = 0; $i < $this->numFactors; ++$i) { |
| 54 | $factors[$i] = \mt_rand() / \mt_getrandmax(); |
| 55 | } |
| 56 | |
| 57 | return $factors; |
| 58 | } |
| 59 | |
| 60 | public function predict($userId, $itemId) { |
| 61 | $userFactor = $this->userFactors[$userId]; |
| 62 | $itemFactor = $this->itemFactors[$itemId]; |
| 63 | $score = 0; |
| 64 | |
| 65 | for ($i = 0; $i < $this->numFactors; ++$i) { |
| 66 | $score += $userFactor[$i] * $itemFactor[$i]; |
| 67 | } |
| 68 | |
| 69 | return $score; |
| 70 | } |
| 71 | |
| 72 | public function updateFactors($userId, $posItemId, $negItemId) : void |
| 73 | { |
| 74 | if (!isset($this->userFactors[$userId])) { |
| 75 | $this->userFactors[$userId] = $this->generateRandomFactors(); |
| 76 | } |
| 77 | |
| 78 | if (!isset($this->itemFactors[$posItemId])) { |
| 79 | $this->itemFactors[$posItemId] = $this->generateRandomFactors(); |
| 80 | } |
| 81 | |
| 82 | if (!isset($this->itemFactors[$negItemId])) { |
| 83 | $this->itemFactors[$negItemId] = $this->generateRandomFactors(); |
| 84 | } |
| 85 | |
| 86 | $userFactor = $this->userFactors[$userId]; |
| 87 | $posItemFactor = $this->itemFactors[$posItemId]; |
| 88 | $negItemFactor = $this->itemFactors[$negItemId]; |
| 89 | |
| 90 | for ($i = 0; $i < $this->numFactors; ++$i) { |
| 91 | $userFactor[$i] += $this->learningRate * ($posItemFactor[$i] - $negItemFactor[$i]) - $this->regularization * $userFactor[$i]; |
| 92 | $posItemFactor[$i] += $this->learningRate * $userFactor[$i] - $this->regularization * $posItemFactor[$i]; |
| 93 | $negItemFactor[$i] += $this->learningRate * (-$userFactor[$i]) - $this->regularization * $negItemFactor[$i]; |
| 94 | } |
| 95 | |
| 96 | $this->userFactors[$userId] = $userFactor; |
| 97 | $this->itemFactors[$posItemId] = $posItemFactor; |
| 98 | $this->itemFactors[$negItemId] = $negItemFactor; |
| 99 | } |
| 100 | } |