Title: Filter-based feature selection: a comparison among binary and continuous Cuckoo optimisation algorithms along with multi-objective optimisation algorithms using gain ratio-based entropy

Authors: Ali Muhammad Usman; Umi Kalsom Yusof; Syibrah Naim

Addresses: Department of Computer Sciences, Federal College of Education (Technical), Gombe, Nigeria ' School of Computer Sciences, Universiti Sains Malaysia, 11800 Pulau Pinang, Malaysia ' Computer and Information Science, University of Strathclyde, Glasgow, UK

Abstract: Filter-based feature selection used the intrinsic statistical characteristic to select high-rank features from a dataset. However, it affects the classification performance due to lack of feature interaction among the selected features. Gain ration (GR)-based entropy which is a modification of the commonly used information gain (IG) is employed as a filter-based evaluation measure on cuckoo optimisation algorithm (COA) along with its binary counterpart (BCOA) together with non-dominated sorting genetic algorithm (NSGA-III), multi-objective evolutionary algorithm (MOEA) based on decomposition and evolutionary algorithm of non-dominated sorting with radial basis (ENORA). The results achieved showed the superiority of the proposed entropy with GR over the existing entropy with IG in most of the datasets. While BCOA perform better than COA in majority of the datasets. The proposed multi-objective algorithms perform better than both BCOA and COA on the majority of the datasets. NSGA-III performed better than all on majority of the datasets.

Keywords: multi-objective optimisation; Cuckoo optimisation algorithm; COA; feature selection; entropy; information gain; gain ration; NSGA-III; ENORA; MOEA/D.

DOI: 10.1504/IJBIC.2022.127511

International Journal of Bio-Inspired Computation, 2022 Vol.20 No.3, pp.183 - 192

Accepted: 12 Oct 2021
Published online: 07 Dec 2022 *

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