Research Article
A Many-Objective Squirrel Hybrid Optimization Algorithm: MaSHOA
@INPROCEEDINGS{10.1007/978-3-030-72792-5_36, author={Zhuoran Liu and Fanhao Zhang and Xinyuan Wang and Qidong Zhao and Changsheng Zhang and Bin Zhang}, title={A Many-Objective Squirrel Hybrid Optimization Algorithm: MaSHOA}, proceedings={Simulation Tools and Techniques. 12th EAI International Conference, SIMUtools 2020, Guiyang, China, August 28-29, 2020, Proceedings, Part I}, proceedings_a={SIMUTOOLS}, year={2021}, month={4}, keywords={Many-objective optimization Squirrel search algorithm Adjustable reference points strategy}, doi={10.1007/978-3-030-72792-5_36} }- Zhuoran Liu
Fanhao Zhang
Xinyuan Wang
Qidong Zhao
Changsheng Zhang
Bin Zhang
Year: 2021
A Many-Objective Squirrel Hybrid Optimization Algorithm: MaSHOA
SIMUTOOLS
Springer
DOI: 10.1007/978-3-030-72792-5_36
Abstract
Many-objective optimization problems (MaOP) are important to the field of computing intelligence which leads to more requirements for the evolutionary many-objective Algorithms (EMaOA). Meanwhile, we consider that the evolution process also has some influence on the performance of results. And we present a many-objective squirrel hybrid optimization algorithm (MaSHOA) which takes an effective squirrel search algorithm (SSA) as the evolution framework and a reference-point-based many-objective evolutionary algorithm (NSGA-III) as the EMaOA framework. This paper applies the scalarizing evaluation to make sure the solution quality among the neighborhood and takes the reference point association achievement as the reference-point-based part. Taking iterations into account, we design a joint fitness function. For both the evolution and selection operations, a joint fitness function is applied to sort solutions to guide others and select them respectively. Besides, the distance penalization is introduced to prevent the local convergence. About useless reference points, this paper proposes an adjustable reference points strategy. The simulation experiment of the proposed algorithm is carried on different test problems with 3 to 15 objectives. Compared with other classic EMaOAs, the means, variances, box plots and parallel coordinate plots of the obtained results are utilized to analyze the convergence and diversity. And this proposed algorithm has good performance on solving MaOPs.
