Abstract: This paper discussed the problems in text categorization accuracy. In traditional text classification algorithm,different feature words have the same affecte on classification result, and classification accuracy is lower, causing the increase algorithm time complexity. Because the maximum entropy model can integrated various relevant or irrelevant probability knowledge observed, the processing of many issues can achieve better results. In order to solve the above problems, this paper proposed an improved maximum entropy text classification, which fully combines rmcan and maximum entropy algorithm advantages. I}he algorithm firstly takes Shannon entropy as maximum entropy model of the objective function, simplifies classifier expression form, and then uses c-mean algorithm to classify the optimal feature. The simulation results show that the proposed method can quickly get the optimal classification feature subsets,grcatly improve text classification accuracy, compared with the traditional text classification.

Key words: Next classification, Maximum entropy algorithm, C-mean, Feature selection