New Feature Selection Process to Enhance Naïve Bayes Classification

Naive Bayes is one of the states of art classification algorithm for data mining applications. Numerous classification techniques have been implemented using Naïve Bayes in the past works. The widely used Mutual Information for feature selection is an empirical method. In this paper, we have used three more variations of mutual information as feature selection methods. The performance of simple and robust Naïve Bayes algorithm is enhanced by the new feature selection method. The benchmark datasets of UCI repository are used to prove the robustness of the feature selection method.