feature_selection

= Feature Selection

Feature Selection is a library of feature selection algorithms.

http://en.wikipedia.org/wiki/Feature_selection

== Install

gem sources -a http://gemcutter.org sudo gem install feature_selection

== How To Use

There are currently 3 implemented feature collections: Chi-Squared, Mutual Information and Frequency Based.

Each return a hash that looks similar to:

{klass => {term => score, term => score}, klass => {term => score}}

Example:

data = { :spam => [['this', 'is', 'some', 'information'], ['this', 'is', 'something', 'that', 'is', 'information']], :ham => [['this', 'test', 'some', 'more', 'information'], ['there', 'are', 'some', 'things']], }

a = FeatureSelection::ChiSquared.new(data)

You can also use...

FeatureSelection::MutualInformation

FeatureSelection::FrequencyBased

a.rank_features #=> {:spam => {term => score, term => score}, :ham => {term => score}}

== Logging

There are two ways to log the activity:

Provide a path of somewhere to log to

log = File.expand_path(File.dirname(FILE) + '/log.txt') FeatureSelection::MutualInformation.new(data, :log_to => log)

Provide an existing Logger object

log = Logger.new('log.txt') FeatureSelection::MutualInformation.new(data, :log_to => log)

== Copyright

Copyright (c) 2009 reddavis. See LICENSE for details.