Feature selection method based on the improved of mutual information and genetic algorithm

Abstract

The feature selection is a key method of text categorization technology, this paper proposed a text feature selection method based on the improved of mutual information and genetic algorithm. Used the improved of mutual information algorithm to do the initial choose to removing redundancy and noise words at first, and then used the genetic algorithm to training the template which generate by a subset of words, so get the optimal feature subset that on behalf of the issue space, to achieve dimensionality reduction and improved classification accuracy.