Research Area: Metaheuristic Computing
This paper focuses on the performance of an improved algorithm based on the salp swarm algorithm (SSA), called AGSSA. We planned several new ideas to improve the defects of the original optimizer, such as ease to fall into local optimum and low convergence accuracy. To solve these problems, the SSA algorithm is improved in two parts. Salp swarm algorithm (SSA) is a recently proposed optimization algorithm with advantages and disadvantages, simulating a perception of the salps foraging and navigation behavior in the deep ocean. The first improvement includes the adaptive control parameter introduced into the follower position update stage, which boosts the local exploitative ability of the population. The second improvement includes the elite gray wolf domination strategy introduced in the last stage of the population position update, which helps the population find the globally optimal solution faster. The performance of AGSSA is verified by a series of problems, including the IEEE CEC2014 benchmark functions, engineering design problems, and feature selection tasks. The experimental results of AGSSA are compared with some well-known metaheuristic algorithms. Simulations reveal that the performance of AGSSA is significantly better than lots of competitive metaheuristic algorithms. Moreover, in solving real-world problems, AGSSA also shows high accuracy in comparison with other metaheuristic algorithms. These points prove that the introduction of the two strategies has a positive effect on the original SSA. Promisingly, the proposed AGSSA can be used as a potential optimization tool in many optimization tasks.
Keywords:
Salp swarm algorithm
Swarm intelligence
Engineering design
Global optimization
Feature selection
Author(s) Name: Songwei Zhao, Pengjun Wang,Xuehua Zhao,Hamza Turabieh,Majdi Mafarja,Huiling Chen
Journal name: Springer
Conferrence name:
Publisher name: Springer
DOI: 10.1007/s00366-021-01464-x
Volume Information: 4501–4528
