Research on text classification mining based on Naive Bayes

In the method of text classification mining based on Naive Bayes, every attribute of eigenvector is independent relatively to confirm a class variable is supposed. But the hypothesis is always disobedient with actual things. So the model of CLIF-NB text classification learning based on Naive Bayes was proposed in the paper The method uses the theory of mutual information to calculate the maximum relative probability in feature attributes of training texts, and introduces variables sets to combine and replace line inseparable attributes to relax limit of independence hypothesis of Naive Bayes.