Yab, Li Yu and Wahid, Noorhaniza and A Hamid, Rahayu (2024) Impact of Balanced Exploration and Exploitation on High-dimensional Feature Selection with Hierarchical Whale Optimisation Algorithm. Journal of ICT, 23 (4). pp. 593-626. ISSN 1675-414X

Preview

PDF - Published Version Available under License Attribution 4.0 International (CC BY 4.0). Download (1MB) | Preview

Abstract

High-dimensional datasets (HDDs) pose significant challenges in feature selection due to their complex nature. While metaheuristic algorithms like the Whale Optimisation Algorithm (WOA) have shown promise in addressing these challenges, stability issues are least addressed. Through an extensive literature review, it was identified that stability is manifested in the equilibrium of exploration and exploitation. Hence, this study introduced the Hierarchical Whale Optimisation Algorithm (HiWOA), an approach designed to enhance the WOA’s stability and performance in HDD feature selection tasks. The HiWOA incorporates a two-phase strategy comprising a nonlinear control parameter based on the arcsine function and a hierarchical position-update mechanism adapted from the Grey Wolf Optimiser. The proposed HiWOA was evaluated through 23 benchmark optimisation functions and feature selection experiments on 11 medical HDDs. The results indicate that the HiWOA outperformed the WOA and a modified variant (mWOA) in terms of a better fitness value, a more balanced exploration-exploitation ratio, and improved classification accuracy with fewer selected features. These findings demonstrate the HiWOA’s effectiveness in enhancing stability, making it a robust solution for high-dimensional optimisation and feature selection

Item Type:	Article
Uncontrolled Keywords:	Balancing exploration and exploitation, control parameter, feature selection, stability, whale optimisation algorithm
Subjects:	Q Science > QA Mathematics
Divisions:	School of Computing
Depositing User:	Mdm. Rozana Zakaria
Date Deposited:	30 Apr 2025 07:00
Last Modified:	30 Apr 2025 07:00
URI:	https://repo.uum.edu.my/id/eprint/32222

Actions (login required)

View Item