Improved Feature Selection Based on Normalized Mutual Information

Abstract

For the question (NMIFS) algorithm has the disadvantages of redundancy. This paper introduces a new feature selection method by enhanced NMIFS algorithm. A new quality estimation function is introduced in the new feature selection algorithm to overcome the shortcomings of the classic NMIFS, and the experiment shows on that normalized mutual information feature selection The experiment shows that the INMIFS can generate impressive results in accuracy and redundancy.