Opposition Based Binary Particle Swarm Optimization Algorithm for Feature Selection

Abstract

In this study, we propose a Binary Particle Swarm Optimization algorithm hybridizing with Oppositionbased Learning for solving the feature selection problem. Opposition-based Learning is used in three different ways: (1) opposition-based population initialization; (2) opposition-based generation jumping; and (3) opposition-based population initialization and generation jumping. We conduct experiments on two medicine data sets. Based on the results, the oppositionbased population initialization and generation jumping performs better. Additionally, we investigate the effect of the eight different transfer functions on the performance of the proposed approach. Among the eight transfer functions, the sigmoid function (S1(x)) yields better performance than others. To evaluate the performance of the proposed method, the Binary Particle Swarm Optimization algorithm is applied to the problem. The results reveal that our approach outperforms the other methods.