Abstract
Feature selection is one of the most important issues in improving the data classification process. It greatly influences the accuracy of the classification. There are many evolutionary algorithms used for this purpose, such as the Particle Swarm Optimization (PSO) in discrete space through the Binary PSO concept. The BPSO optimization algorithm derives its mechanism from the default PSO algorithm but in discrete space. In this research, a hybrid approach was proposed between the BPSO algorithm and Mutual Information (MI) to obtain subsets of features through two basic phases: the first is to use the BPSO algorithm to determine the features affecting the data classification process by relying on an objective function. In the second phase, the MI method is used to reduce the number of features identified by the BPSO method. The results of the proposed algorithm have demonstrated efficiency and effectiveness by obtaining higher classification accuracy and using fewer features than default methods.
