camspiers/statistical-classifier

A PHP implementation of Complement Naive Bayes and SVM statistical classifiers, including a structure for building other classifier, multiple data sources and multiple caching backends

PHP Classifier

, _(*1)

PHP Classifier uses semantic versioning, it is currently at major version 0, so the public API should not be considered stable., _(*2)

What is it?

PHP Classifier is a text classification library with a focus on reuse, customizability and performance. Classifiers can be used for many purposes, but are particularly useful in detecting spam., _(*3)

Features

• Complement Naive Bayes Classifier
• SVM (libsvm) Classifier
• Highly customizable (easily modify or build your own classifier)
• Command-line interface via separate library (phar archive)
• Multiple data import types to get your data into the classifier (Directory of files, Database queries, Json, Serialized arrays)
• Multiple types of model caching
• Compatible with HipHop VM

Installation

$ composer require camspiers/statistical-classifier

SVM Support

For SVM Support both libsvm and php-svm are required. For installation intructions refer to php-svm., _(*4)

Usage

Non-cached Naive Bayes

use Camspiers\StatisticalClassifier\Classifier\ComplementNaiveBayes;
use Camspiers\StatisticalClassifier\DataSource\DataArray;

$source = new DataArray();
$source->addDocument('spam', 'Some spam document');
$source->addDocument('spam', 'Another spam document');
$source->addDocument('ham', 'Some ham document');
$source->addDocument('ham', 'Another ham document');

$classifier = new ComplementNaiveBayes($source);
$classifier->is('ham', 'Some ham document'); // bool(true)
$classifier->classify('Some ham document'); // string "ham"

Non-cached SVM

use Camspiers\StatisticalClassifier\Classifier\SVM;
use Camspiers\StatisticalClassifier\DataSource\DataArray;

$source = new DataArray()
$source->addDocument('spam', 'Some spam document');
$source->addDocument('spam', 'Another spam document');
$source->addDocument('ham', 'Some ham document');
$source->addDocument('ham', 'Another ham document');

$classifier = new SVM($source);
$classifier->is('ham', 'Some ham document'); // bool(true)
$classifier->classify('Some ham document'); // string "ham"

Caching models

Caching models requires maximebf/CacheCache which can be installed via packagist. Additional caching systems can be easily integrated., _(*5)

Cached Naive Bayes

use Camspiers\StatisticalClassifier\Classifier\ComplementNaiveBayes;
use Camspiers\StatisticalClassifier\Model\CachedModel;
use Camspiers\StatisticalClassifier\DataSource\DataArray;

$source = new DataArray();
$source->addDocument('spam', 'Some spam document');
$source->addDocument('spam', 'Another spam document');
$source->addDocument('ham', 'Some ham document');
$source->addDocument('ham', 'Another ham document');

$model = new CachedModel(
    'mycachename',
    new CacheCache\Cache(
        new CacheCache\Backends\File(
            array(
                'dir' => __DIR__
            )
        )
    )
);

$classifier = new ComplementNaiveBayes($source, $model);
$classifier->is('ham', 'Some ham document'); // bool(true)
$classifier->classify('Some ham document'); // string "ham"

Cached SVM

use Camspiers\StatisticalClassifier\Classifier\SVM;
use Camspiers\StatisticalClassifier\Model\SVMCachedModel;
use Camspiers\StatisticalClassifier\DataSource\DataArray;

$source = new DataArray();
$source->addDocument('spam', 'Some spam document');
$source->addDocument('spam', 'Another spam document');
$source->addDocument('ham', 'Some ham document');
$source->addDocument('ham', 'Another ham document');

$model = new Model\SVMCachedModel(
    __DIR__ . '/model.svm',
    new CacheCache\Cache(
        new CacheCache\Backends\File(
            array(
                'dir' => __DIR__
            )
        )
    )
);

$classifier = new SVM($source, $model);
$classifier->is('ham', 'Some ham document'); // bool(true)
$classifier->classify('Some ham document'); // string "ham"

Unit testing

statistical-classifier/ $ composer install --dev
statistical-classifier/ $ phpunit