aldeebhasan/laravelcf

This package will allow you to make fast recommendation based on custom inputs

1.0.3 2023-07-10 17:17 UTC

Requires

• php: >=8.1
• ext-json: *

Requires (Dev)

• orchestra/testbench: 8.x-dev
• phpunit/phpunit: ^9.6

MIT

• Hasan Deeb

• laravel
• statistics
• recommendation
• collaborative filtering
• fast recommendation
• product recommendations

README

A php package allow you to find the est recommendation of your modules

Installation

Install using composer:

composer require aldeebhasan/laravelcf

The run :

php artisan migrate

Basic Usage

LarvelCF package allow you to recommend data based on many algorithms including (Cosine,Weighted Cosine, Centered Cosine, SlopeOne).

In general we have two kind of recommender included in this package:

• item-based recommender
• user-based recommender

Filling the data

The first step to build your recommendation system is to provide the dataset you want to work with.

In LarvelCF we have 4 type of data (PURCHASE, RATE, CART_ACTION, BOOKMARK). the purpose of these types is to enable you to handle different types of data at the same time.

You can use the following to enter your recommender data:

use \Aldeebhasan\LaravelCF\Facades\Recommender;
  
Recommender::addRating      ('user_1', 'product_1', 5);
Recommender::addCartAddition('user_1', 'product_1', 2); // like the quantity
Recommender::addPurchase    ('user_1', 'product_1', 5); // like the quantity
Recommender::addBookmark    ('user_1', 'product_1', 5);

Instantiate the recommender

After entering your data, you can instantiate your desired recommender using our support facade:

use \Aldeebhasan\LaravelCF\Facades\Recommender;
use \Aldeebhasan\LaravelCF\Enums\RelationType;

/* you cann also use any of RelationType::PURCHASE,RelationType::CART_ACTION,RelationType::BOOKMARK*/
Recommender::getItemBasedRecommender(RelationType::RATE); // to recommend similar products
//OR
Recommender::getUserBasedRecommender(RelationType::RATE);// to recommend similar users

Get recommendations

Finally, to make your recommendations you will run the following code:

use \Aldeebhasan\LaravelCF\Facades\Recommender;
use \Aldeebhasan\LaravelCF\Enums\RelationType;

/* you cann also use any of RelationType::PURCHASE,RelationType::CART_ACTION,RelationType::BOOKMARK*/
Recommender::getItemBasedRecommender(RelationType::RATE)
            ->setSimilarityFunction(Cosine::class)
            ->train()
            ->recommendTo('user_1');

For the setSimilarityFunction, you can provide the similarity algorithm, the missing value default values, and weather you want to fill the missing methods or discard them.

Available Similarity algorithm:

• Cosine::class (default for item-based)
• CosineCentered::class
• CosineWeighted::class
• Jaccard::class
• SlopeOne::class
• Pearson::class (default for user-based)

Available missing values replacement methods:

• MissingValue::ZERO (package default)
• MissingValue::MEAN
• MissingValue::MEDIAN

use \Aldeebhasan\LaravelCF\Facades\Recommender;
use \Aldeebhasan\LaravelCF\Enums\RelationType;
use \Aldeebhasan\LaravelCF\Enums\MissingValue;
use \Aldeebhasan\LaravelCF\Similarity;

Recommender::getItemBasedRecommender(RelationType::RATE)
            // use Weighted cosine algorithm and replace the missing values with zero 
            ->setSimilarityFunction(CosineWeighted::class,MissingValue::ZERO,true)
             // use SlopeOne algorithm and replace the missing values with the mean 
            ->setSimilarityFunction(SlopeOne::class,MissingValue::MEAN,true)
            

Full Example

use \Aldeebhasan\LaravelCF\Facades\Recommender;
use \Aldeebhasan\LaravelCF\Enums\RelationType;
use \Aldeebhasan\LaravelCF\Enums\MissingValue;
use \Aldeebhasan\LaravelCF\Similarity;

Recommender::addRating(1, 'squid', 1);
Recommender::addRating(2, 'squid', 1);
Recommender::addRating(3, 'squid', 0.2);
Recommender::addRating(1, 'cuttlefish', 0.5);
Recommender::addRating(3, 'cuttlefish', 0.4);
Recommender::addRating(4, 'cuttlefish', 0.9);
Recommender::addRating(1, 'octopus', 0.2);
Recommender::addRating(2, 'octopus', 0.5);
Recommender::addRating(3, 'octopus', 1);
Recommender::addRating(4, 'octopus', 0.4);
Recommender::addRating(1, 'nautilus', 0.2);
Recommender::addRating(3, 'nautilus', 0.4);
Recommender::addRating(4, 'nautilus', 0.5);
$results = Recommender::getItemBasedRecommender(RelationType::RATE)
    ->setSimilarityFunction(CosineWeighted::class, MissingValue::MEAN)
    ->train()
    ->recommendTo('squid');


/**
recommendation results sorted by similarity: 
[ 
  "cuttlefish" => 0.89
  "nautilus" => 0.75
  "octopus" => 0.5
]
**/

License

Laravel Recommendation system package is licensed under The MIT License (MIT).

Security contact information

To report a security vulnerability, contact directly to the developer contact email Here.