An Enhanced Zebra Optimization Algorithm With Multiple Strategies for Global Optimization and Feature Selection Problems: A Hepatocellular Carcinoma Case Study

The Zebra Optimization Algorithm (ZOA) is a newly proposed biology-based metaheuristic method drawn from zebras in nature, designed to tackle optimization problems. As with other metaheuristic algorithms, ZOA has limitations, including a decrease in population diversity, an imbalance between exploration and exploitation, restricted exploration abilities, and the inclination to get stuck in local optima. To tackle them, this paper proposes an enhanced version of ZOA with multiple strategies called CLESQ-ZOA, which combines three strategies: chaotic mapping, Logarithmic Spiral Strategy (LSS), and Enhanced Search Quality (ESQ). In CLESQ-ZOA, the initial population is produced utilizing chaotic mapping to increase population diversity and provide a more uniform and diverse spread of the initial solutions. The integration of LSS with CLESQ-ZOA allows for greater search capabilities and more efficient exploration of the search space by each individual. Moreover, the ESQ is integrated into the algorithm to minimize the likelihood of getting trapped in local optima by maintaining an equilibrium between exploration and exploitation. These multiple strategies are implemented to enhance population diversity, boost the search capability of CLESQ-ZOA, and strike a delicate balance between exploration and exploitation. The effectiveness of CLESQ-ZOA has been evaluated using standard benchmark functions. Additionally, a binary version of CLESQ-ZOA is also proposed in this paper. The binary version of CLESQ-ZOA is applied to twelve feature selection benchmark datasets to demonstrate its effectiveness in real-world applications. Besides, CLESQ-ZOA has been applied to Hepatocellular Carcinoma to further evaluate its effectiveness. The findings obtained from the CLESQ-ZOA are compared with various metaheuristic methods and the performance is analyzed with Wilcoxon rank-sum and Friedman tests. Experimental and statistical results confirm the effectiveness and reliability of CLESQ-ZOA. Evaluations of feature selection datasets show that CLESQ-ZOA exhibits effective performance in determining the optimal feature subset.