Title: New transformation method in continuous particle swarm optimisation for feature selection

Authors: Kangshun Li; Dunmin Chen; Zhaolian Zeng; Guang Chen; James Tin-Yau Kwok

Addresses: School of Mathematics and Informatics, South China Agricultural University, Guangzhou, Guangdong, China; School of Computer Science, Guangdong University of Science and Technology, Dongguan, Guangdong, China ' School of Mathematics and Informatics, South China Agricultural University, Guangzhou, Guangdong, China ' School of Foreign Studies, South China Agricultural University, Guangzhou, Guangdong, China ' GRG Banking Equipment Co., Ltd., Guangzhou, China ' Department of Computer Science and Engineering, Hong Kong University of Science and Technology, Sai Kung, New Territories, Hong Kong

Abstract: Feature selection is a very important task in many real-world problems. Because of its powerful search ability, Particle Swarm Optimisation (PSO) is widely applied to feature selection. However, PSO was originally designed for continuous problems, and therefore, the transformation between continuous particles and binary solutions is needed. This paper proposes a new transformation methods-based PSO (PSOS) in which the related feature subset of a particle is decided by a sine function rather than comparing with a single threshold. To further upgrade the performance of the proposed method, an extra increment generated by the Gaussian distribution is added to the marginal positions (PSOSI). The experimental results show that PSOS and PSOSI can select smaller feature subsets with higher classification accuracy than all the other algorithms compared in this paper. Furthermore, in most cases, the performance of the second method is better than the first one.

Keywords: PSO; particle swarm optimisation; feature selection; classification; sine function; Gaussian distribution; transformation method.

DOI: 10.1504/IJWMC.2022.123290

International Journal of Wireless and Mobile Computing, 2022 Vol.22 No.2, pp.114 - 124

Received: 05 Jun 2020
Accepted: 08 Sep 2020
Published online: 08 Jun 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article