Incremental Learning Of Naive Bayes Chinese Classification System

With the rapid development of Internet, text information increases exponentially. In order to control and retrieve valuable information, research of automatic text categorization becomes very important. Naive Bayes method is a simple and effective probability categorization method at present, but this algorithm has the lack of non-incremental learning ability. This paper proposes an incremental learning algorithm based on weighted naive bayes to deal with this lack. It introduces the main idea of incremental learning and the weighted naive Baye cremental learning algorithm in detail, and designes a simple incremental Bayes Chinese text classification experimental system.The content of this thesis mainly includes the following aspects:1. This paper describes the general process of text categorization. It introduces a few Bayesain classification models based on Bayesain theory. It analyzes and compares their individual characteristics.2. This the paper proposes a new TFIDF feature selection method with concept of information entropy based on a thorough study of lots of feature selection methods. Experimental results show the features selected by this method has more representation than before.3. This paper proposes an incremental learning algorithm based on weighted na(?)ve Bayes to deal with this lack of non-incremental learning for traditional naive Bayes. It gives the basic idea of the very algorithm. It proposes selective incremental learning by setting up class confidence threshold. It proposes the incremental learning algorithm based on weighted naive Bayes and gives a detailed proof and analysis.4. This paper designes a simple incremental Bayes Chinese text classification experimental system. It does experiments using two Chinese data sets. Experimental results show that the effect of incremental Bayes is better than non-incremental Bayesian.