Naive Bayesian Classifier Using Feature Weighting Based on Rough Sets

Naive Bayesian classifier is a simple and efficient classification algorithm,but its attribute independence assumption affects the classification results.Relaxing "Naive Bayes assumption" can enhance the effect of naive Bayesian classification and usually result in a substantial increase in the computational cost.In this paper,a Nave Bayesian Classifiers using feature weighting based on rough sets is proposed.The weighted parameters learn directly from the training data and are regarded as the significance of each attribute to the particular class value when evaluating the posterior probability.In comparison the classification algorithms with the naive Nave Bayesian classifier,Bayesian Networks and NBTree,experimental results show that FWNB classifier has a higher classification accuracy at the cost of less computation than others in most data sets.