Balanced Exploration and Exploitation Properties of Multi-Objective Sooty Tern Optimizer

EasyChair Preprint 11569

Abstract

The Sooty Tern Optimization Algorithm (STOA), a newly designed metaheuristic drawing inspiration from the migratory and predatory behaviors of the sea bird Sooty Tern, presents several notable advantages, including minimal parameterization. However, it is confined to addressing singleobjective problems exclusively. In this work, we have enhanced the properties of exploration and exploitation to effectively penetrate the area of search. Subsequently, we extended the STOA to a multi-objective version called MOSTOA (MultiObjective Sooty Tern Optimization Algorithm). This enhanced algorithm is designed to address multiple objectives in diverse problem domains. The MOSTOA utilizes an archive repository to store and retrieve the optimal solutions generated throughout the optimization cycle. From this population archive, leaders are chosen to guide the solutions of the main population towards promising search locations. Furthermore, the utilization of the grid mechanism and dynamic archiving approach serves the purpose of achieving a harmonious equilibrium between convergence and variety inside the final Pareto set. These strategies ensure that the obtained solutions exhibit both high quality and spread across the objective space. The proposed MOSTOA is validated on various well-known benchmarks functions. In addition, its performance is assessed in comparison to well-established cutting-edge algorithms. Our method produces very competitive results and, in most circumstances, exhibits improved convergence behavior with a good variety of solutions, as demonstrated by the experimental findings.

Keyphrases: Metaheuristic, Pareto front, Swarm Intelligence, dominance relation, multi-objective optimization

@booklet{EasyChair:11569,
  author    = {Farid Boumaza and Djaafar Zouache and Abou El Hassane Benyamina},
  title     = {Balanced Exploration and Exploitation Properties of Multi-Objective Sooty Tern Optimizer},
  howpublished = {EasyChair Preprint 11569},
  year      = {EasyChair, 2023}}