Text categorization based on granular agent evolutionary classification algorithm

Document classification, with the blooming of the Internet information delivery, has become indispensable required and is expected to be disposed by an automatic text categorization. This paper presents a text categorization approach based on granular agent evolutionary classification algorithm to the single-labeled documents. First, a new feature selection method combined term frequency with class information is proposed according to the analyses of existed approaches. It based on the term weighting scheme, and using some useful information in other feature selection. Second, inspiration of the ideas in granular agent evolutionary classification algorithm, a new classifier is introduced in the classifying module. It causes the evolution of sets of documents, and at the end of the evolutionary process, extracts rules from these sets. Because the particularity in text categorization, some specific operators are devised for realizing the evolutionary operations performed on granular agent. Assimilation operator, exchange operator, and differentiation operator reflect the competitive, cooperative and self-learning ability of agent respectively. In experiments, the effectiveness of the proposed approach is evaluated in Reuters-21578. The test results show that the algorithm has a good recall, precision and F1 measure. In most categories, the performance of it is better than Naïve Bayes, K-nearest neighbor and support vector machine, which have good performance on the text categorization. All the results show the proposed algorithm is good. © 2016 American Scientific Publishers All rights reserved.