Intelligent Feature Selection by Bacterial Foraging Algorithm and Information Theory

Abstract

In this paper, an intelligent feature selection by bacterial foraging algorithm and mutual information is proposed. Feature selection is an important issue in the pattern classification problem. Particularly, in the case of classifying with a large number of features or variables, the accuracy and computational time of the classifier can be improved by using the relevant feature subset to remove the irrelevant, redundant, or noisy data. The proposed method consists of two parts: a wrapper part by bacterial foraging optimization and a filter part by mutual information. In order to select the best feature subset to achieve the best performance of the classifiers. Experimental results show that this method can achieve better performance for pattern recognition problems other than other conventional ones.