This is an implementation of a Naive Bayes classifier in CoffeeScript.

Dependencies

This project will use PorterStemmer1980.js (included as a git submodule in this project) if it finds the stemmer variable in the global namespace, otherwise stemming is not performed.

Implementation

Documents have stop-words removed and are stemmed before training and classifying.

Usage

###Basic usage

c = new NaiveBayesClassifier()
c.train("Something interesting to you", "mundane sentence")
klass = c.classify("another phrase")

Other Features

Saving a trained classifier

Call the to_object function to get an object representing the internal data of the classifier. The object can be saved and passed to the constructor to re-constitute the classifier at another time.

save = c.to_object()
c2 = new NaiveBayesClassifier(save)

Untraining

Call the untrain method with a document and class to untrain it. This might be useful for applications that use a classifier but need to support undo.

c = new NaiveBayesClassifier()
c.untrain("Something interesting to you", "mundane sentence")

Renaming a class

You can rename a class in the classifier. For example, you were previously training documents with the class name of "phyics" and realized your spelling mistake, you could do the following:

c.rename_class('phyics', 'physics')