Scholarlink

<p>To address the issues of slow convergence and susceptibility to local optima in the coati optimization algorithm (COA), this study proposes a multi-strategy enhanced variant named WOCOA. WOCOA incorporates a hybrid initialization strategy combining random initialization and lens opposition-based learning (LOBL), further enhanced by Tent mapping to increase population diversity. The golden sine strategy is introduced in the search phase to balance exploration and exploitation, while a spiral bubble-net strategy from whale optimization algorithm (WOA) is applied in the exploitation phase to enhance global search capability. Ablation experiments validate the individual and collective contributions of these strategies to performance improvement. Furthermore, comparative experiments across 18 UCI datasets, 23 benchmark test functions, and the CEC-2017 suite demonstrate the effectiveness and superiority of WOCOA in feature selection and numerical optimization. Finally, two scalability experiments confirm its applicability and robustness in solving high-dimensional and complex real-world optimization problems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-strategy improved coati optimization algorithm

  • 期刊论文

  • 发表日期：

    2025年7月18日

查看全文
  • Yujie Xu,
  • Yao Jiang,
  • Jialing Xu,
  • Lan Wang,
  • Xingchen Liu,
  • Liyun Jia

摘要

To address the issues of slow convergence and susceptibility to local optima in the coati optimization algorithm (COA), this study proposes a multi-strategy enhanced variant named WOCOA. WOCOA incorporates a hybrid initialization strategy combining random initialization and lens opposition-based learning (LOBL), further enhanced by Tent mapping to increase population diversity. The golden sine strategy is introduced in the search phase to balance exploration and exploitation, while a spiral bubble-net strategy from whale optimization algorithm (WOA) is applied in the exploitation phase to enhance global search capability. Ablation experiments validate the individual and collective contributions of these strategies to performance improvement. Furthermore, comparative experiments across 18 UCI datasets, 23 benchmark test functions, and the CEC-2017 suite demonstrate the effectiveness and superiority of WOCOA in feature selection and numerical optimization. Finally, two scalability experiments confirm its applicability and robustness in solving high-dimensional and complex real-world optimization problems.