Text classification based on the TAN model

Abstract

This paper proposes a text classification method based on TAN model. Naive Bayesian classifier is the most effective and popular text classification method, but its attribute independence assumption makes it unable to express the dependence among text terms. TAN (Tree Augmented Naive Bayes) combines the simplicity of Naive Bayesian with the ability to express the dependence among attributes in Bayesian network. This paper reviews some existing text methods, introduces TAN model, and applies TAN model to text classification. Naive Bayesian and TAN classifiers are also compared by our experiments. Experimental results show TAN classifier has better performance.