Naïve Bayes is probably the world's most famous classification technology, and just to keep you on your toes, we provide two separate implementations with lots of configurability. One of the most well-known applications of a Naïve Bayes classifier is for spam filtering in an e-mail.
The reason the word naïve is used is that the classifier assumes that words (features) occur independent of one another—this is clearly a naïve assumption, but lots of useful and not-so-useful technologies have been based on the approach. Some notable features of the traditional naïve Bayes include:
Character sequences are converted to bags of tokens with counts. No whitespaces are considered, and the order of the tokens does not matter.
Naïve Bayes classifiers require two or more categories into which input texts are categorized. These categories must be both exhaustive and mutually exclusive. This indicates that a document used for training must only belong to one category.
The math is very simple:...