Hybrid ACO-PSO based approaches for feature selection

Abstract

Feature selection attempts to find the most discriminative information aiming to design an accurate learning system. Feature selection has been the focus of interest for a long time and many works had been done. Recently, the tendency of research in this domain is oriented to the bio-inspired methods. In this paper, we propose hybrid bio-inspired approaches applied to the feature selection problem. The approaches are based on two swarm intelligence methods: ant colony optimization (ACO) and particle swarm optimization (PSO). The performances of these approaches are compared with simple bio-inspired feature selection methods based on ant colony optimization, particle swarm optimization and genetic algorithm. Our experimental results show the efficiency of the proposed approaches in the reduction of selected features number and improvement of classification performance.