期刊
KNOWLEDGE-BASED SYSTEMS
卷 233, 期 -, 页码 -出版社
ELSEVIER
DOI: 10.1016/j.knosys.2021.107543
关键词
Whale optimization algorithm; High dimensional problem; Benchmark function; Friedman's test; Nemenyi multiple comparison tests; Boxplot
The Whale Optimization Algorithm was developed based on the prey-catching characteristics of humpback whales and has been widely used in various disciplines due to its simplicity and efficiency. However, it has been found to have limitations in exploration ability, accuracy, and convergence in high-dimensional optimization problems. This study introduces a new variant with modifications to address these issues and enhance the algorithm's performance by balancing global and local search phases, modifying co-efficient vectors, and introducing random movement. The proposed algorithm demonstrates better performance on higher-dimensional problems compared to the basic Whale Optimization Algorithm and its variants.
Whale optimization algorithm was developed based on the prey-catching characteristics of the humpback whales. Due to its simple structure and efficiency, the researchers employed the algorithm to address numerous disciplines' numerous problems. The profound analysis of the whale optimization algorithm discloses that the algorithm suffers from low exploration ability, lesser accuracy, and early convergence. Additionally, performance of the whale optimization algorithm and most of its variants in high-dimensional optimization problems is not satisfactory. This study proposes a new variant with several modifications to the basic whale optimization algorithm to solve high-dimensional problems. A unique selection parameter is introduced in the whale optimization algorithm to balance the algorithm's global and local search phase. The co-efficient vectors A and C are modified and used effectively to explore and exploit the search region better. In the exploration phase, random movement is allowed to reduce the computational burden of the algorithm. An inertia weight is introduced in the exploitation phase for exhaustive search nearby the best solution. The proposed algorithm evaluates twenty-five benchmark functions using dimensions 100, 500, 1000, and 2000 and compared the results with the whale optimization algorithm and its variants. The estimated outcomes are also compared with seven basic metaheuristic algorithms. Finally, statistical analysis, complexity analysis, and convergence analysis are performed to establish the algorithm's efficacy. All the test result suggests better performance of the proposed algorithm on higher-dimensional problems. (c) 2021 Elsevier B.V. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据