hybridlogic/classifier

A Naive Bayesian classification library for PHP with support for different tokenizers to optimize string classification.

v1.0.2 2013-01-12 13:30 UTC

Requires

• php: >=5.3.0

MIT

• Luke Lanchester

• tokenizer
• statistics
• groups
• bayes
• classifier

README

A Naive Bayesian classification library for PHP with support for different tokenizers to optimize string classification.

0.0 Table of Contents

• Introduction
• Examples
• Running Tests
• Troubleshooting
• Changelog

1.0 Introduction

Naive Bayesian Classification provides a computationally cheap, but reasonably accurate method of determining which "group" a provided string belongs to given a prior training set.

PHP Classifier supports any number of groups to classify strings against and uses logarithmic counting to prevent integer underflow when using larger sets.

2.0 Examples

$tokenizer = new HybridLogic\Classifier\Basic;
$classifier = new HybridLogic\Classifier($tokenizer);

$classifier->train('hot', 'The sun is hot');
$classifier->train('hot', 'It was a warm day in the sun');
$classifier->train('hot', 'This tea is hot!');

$classifier->train('cold', 'This ice is very cold!');
$classifier->train('cold', 'It\'s cold at night');
$classifier->train('cold', 'Ice formed on my at over night');

$groups = $classifier->classify('It was chilly last night');

3.0 Running Tests

phpunit tests

4.0 Troubleshooting

@todo

5.0 Changelog

• [2013-01-10] Initial port from KoBayes