A Novel Fuzzy-Bayesian Classification Method for Automatic Text Categorization

Abstract

Text categorization is mostly required to label the documents automatically with the predefined set of topics. It has been achieved by the large number of advanced machine learning algorithms. In the proposed system, fuzzy rule along with Bayesian classification method is proposed for automatic text categorization using the class-specific features. The proposed method selects the particular feature subset for each class. Then, these class features are applied for the classification. To achieve this, Baggenstoss’s PDF Projection Theorem is followed to reconstruct PDF in raw data space from the class-specific PDF in low-dimensional feature space and build the fuzzy based Bayes classification rule. The noticeable significance of this method is that most feature selection criteria such as information gain and maximum discrimination which can be easily incorporated into the proposed method. The proposed classification performance is evaluated on different datasets and compared with the different feature selection methods. The experimental results illustrate that the effectiveness of the proposed method and further indicates its wide applications in text categorization.