Refined selfish herd optimizer for global optimization problems
Refined selfish herd optimizer for global optimization problems
复制标题
DOI：
10.1016/j.eswa.2019.112838
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
Adiljan Yimit;Koji Iigura;Y. Hagihara
中科院分区：
文献类型：
--
作者：
Adiljan Yimit;Koji Iigura;Y. Hagihara
自私群体优化算法（SHO）是近年来发展起来的一种求解全局优化问题的群优化算法。SHO模仿了广泛观察到的避免捕食风险的自私羊群行为。在SHO中，一组独特的进化算子的启发，捕食者的关系被用来处理优化问题。虽然SHO在寻找最优解时可以提供良好的性能，但该算法存在一些疏忽，导致SHO的开发能力出现问题。此外，SHO算法还存在一些不足，影响了算法的搜索性能和避免陷入局部最优的能力。本文针对这些疏漏和不足，提出了一些改进和修改。为了验证改进后的算法是否能有效地提高SHO算法的性能，本文采用两组基准测试函数将该算法与原SHO算法以及标准粒子群算法、人工蜂群算法、差分进化算法、乌鸦搜索算法等著名算法进行了比较。使用非参数Wilcoxon秩和检验证明了所提出的方法的有效性。实验结果表明，所提出的修改是适当的，以提高原SHO的性能。此外，与其他算法相比，该方法在寻找全局最优解方面也具有竞争力。
The selfish herds optimizer (SHO) is a recently developed swarm optimization algorithm for solving global optimization problems. SHO mimics the widely observed selfish herd behaviors of avoiding predation risks. In SHO, a set of unique evolutionary operators inspired by the prey-predator relationship are used in dealing with optimization problems. Although SHO can provide a good performance when finding an optimal solution, this algorithm has some oversights that cause problems in the exploitation ability of SHO. Additionally, SHO still has some shortcomings that influence the performance of exploration and avoiding stagnation in local optima. In this paper, some refinements and modifications are proposed for these oversights and shortcomings. In order to validate whether the refinements and modifications are appropriate for improving the performance of SHO, two suites of benchmark functions are employed to compare the proposed method with the original SHO and other well-known and recently developed algorithms, such as Standard Particle Swarm Optimization, Artificial Bee Colony algorithm, Differential Evolution, and Crow Search Algorithm, etc. Finally, the efficiency of the proposed method is justified using the nonparametric Wilcoxon rank-sum test. The experimental results show that the proposed modifications are appropriate for improving the performance of the original SHO. In addition, the proposed method can also give competitive results in finding global optima when compared with other algorithms.