harthur / classifier

Bayesian classifier with Redis backend
MIT License
624 stars 65 forks source link

Deprecation notice: This library is no longer actively maintained. Try the natural classifier. It doesn't have a Redis backend, but otherwise works even better.

classifier

classifier is a JavaScript naive Bayesian classifier with backends for Redis and localStorage:

var bayes = new classifier.Bayesian();

bayes.train("cheap replica watches", 'spam');
bayes.train("I don't know if this works on windows", 'not');

var category = bayes.classify("free watches");   // "spam"

The first argument to train() can be a string of text or an array of words, the second argument can be any category name you want.

using in node

If you have node you can install with npm:

npm install classifier

using in the browser

Download the latest classifier.js. In the browser you can only use the localStorage and (default) memory backends.

Redis backend

You can store the classifier state in Redis for persisting and training from multiple sources:

var bayes = new classifier.Bayesian({
  backend: {
    type: 'Redis',
    options: {
      hostname: 'localhost', // default
      port: 6379,            // default
      name: 'emailspam'      // namespace for persisting
    }
  }
});

bayes.train("cheap replica watches", 'spam', function() {
  console.log("trained");
});

bayes.classify("free watches", function(category) {
  console.log("classified in: " + category);
});

JSON

You can serialize and load in the classifier's state with JSON:

var json = bayes.toJSON();

bayes.fromJSON(json);

Other options

Bayesian() takes an options hash that you can define these properties in:

backend

The backend property takes a type which is one of 'Redis', 'localStorage', or 'memory'(default). The backend also has an options hash. The Redis backend takes hostname, port, password, name, db, and error (an error callback) in its options. The localStorage backend takes name for namespacing.

thresholds

Specify the classification thresholds for each category. To classify an item in a category with a threshold of x the probably that item is in the category has to be more than x times the probability that it's in any other category. Default value is 1. A common threshold setting for spam is:

thresholds: {
  spam: 3,
  not: 1
}

default

The default category to throw an item in if it can't be classified in any of the categories. The default value of default is "unclassified".