IMF-PSO: A Particle Swarm Optimization Algorithm for Feature Selection in Classification

Feature selection is an important step in classification. Its goal is to find a set of features that can lead to high classification accuracy with a smaller number of features. This paper addresses feature selection as an optimization problem and solves it by a particle swarm optimization (PSO)-based approach. In the proposed PSO, we adopt three algorithmic components to enhance its performance: feature space adjustment, multi-swarm search, and local-best-guided improvement. We examine the effects of these components using seven data sets from the UCI repository. We also compare our algorithm with two existing algorithms. Experimental results show that the incorporated algorithmic components improve the algorithm performance and our algorithm outperforms the compared algorithms.