Scholarlink

<p>This paper presents an enhanced version of the Parrot Optimizer (PO), termed the Enhanced Parrot Optimizer (EPO), which integrates hierarchical leadership, mimicry-based behavior, and a weighted behavior selection mechanism to improve convergence behavior and robustness in complex optimization landscapes. The proposed EPO is evaluated through comprehensive benchmark tests, including unimodal, multimodal, and non-convex functions, specifically Sphere, Rosenbrock, and Ackley, as well as a real-world business intelligence (BI) portfolio optimization problem. Experimental results show that EPO consistently outperforms the original PO and other established metaheuristic algorithms, including Harris Hawks Optimization (HHO), Whale Optimization Algorithm (WOA), and Remora Optimization Algorithm (ROA). On the Sphere Function, EPO achieved a mean fitness of 47,200 with a low standard deviation of 1600, outperforming HHO, which obtained a mean of 47,000 with a standard deviation of 2500. In the more challenging Rosenbrock Function, EPO reached the best fitness of 8.42E + 07 and a mean of 1.01E + 08, surpassing the other algorithms in both precision and consistency. For the multimodal Ackley Function, EPO reported the lowest mean fitness of 19.60, with a deviation of 0.19, indicating high robustness and stability. Statistical analysis using the Wilcoxon signed-rank test confirmed the superiority of EPO, with p-values below 10⁻⁸ in most comparisons. Despite the algorithmic enhancements, EPO preserved the computational efficiency, with only a 10 to 12 percent increase in runtime compared to the original PO. Moreover, EPO was used to optimize a real-world BI portfolio case study and scored the Maximum portfolio utility, 92.6, and the extreme individual fitness, 94.21, and had the minimum Maximum average running time of 1.24&#xa0;s. These results show two things: first, EPO is useful not only in raising average performance but also in raising the quality and reliability of solutions in both synthetic and pragmatic optimization applications.</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

A Novel Enhancement of the Parrot Optimizer: integrating Hierarchical and Mimicry Behaviors for Improved Performance—Case Study on Business Intelligence Portfolio Optimization

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月6日

查看全文
  • Mohanad A. Deif,
  • Mohamed Elhoseny,
  • Waleed Alomoush,
  • Mohamed A. Hafez,
  • Mohammad Khishe

摘要

This paper presents an enhanced version of the Parrot Optimizer (PO), termed the Enhanced Parrot Optimizer (EPO), which integrates hierarchical leadership, mimicry-based behavior, and a weighted behavior selection mechanism to improve convergence behavior and robustness in complex optimization landscapes. The proposed EPO is evaluated through comprehensive benchmark tests, including unimodal, multimodal, and non-convex functions, specifically Sphere, Rosenbrock, and Ackley, as well as a real-world business intelligence (BI) portfolio optimization problem. Experimental results show that EPO consistently outperforms the original PO and other established metaheuristic algorithms, including Harris Hawks Optimization (HHO), Whale Optimization Algorithm (WOA), and Remora Optimization Algorithm (ROA). On the Sphere Function, EPO achieved a mean fitness of 47,200 with a low standard deviation of 1600, outperforming HHO, which obtained a mean of 47,000 with a standard deviation of 2500. In the more challenging Rosenbrock Function, EPO reached the best fitness of 8.42E + 07 and a mean of 1.01E + 08, surpassing the other algorithms in both precision and consistency. For the multimodal Ackley Function, EPO reported the lowest mean fitness of 19.60, with a deviation of 0.19, indicating high robustness and stability. Statistical analysis using the Wilcoxon signed-rank test confirmed the superiority of EPO, with p-values below 10⁻⁸ in most comparisons. Despite the algorithmic enhancements, EPO preserved the computational efficiency, with only a 10 to 12 percent increase in runtime compared to the original PO. Moreover, EPO was used to optimize a real-world BI portfolio case study and scored the Maximum portfolio utility, 92.6, and the extreme individual fitness, 94.21, and had the minimum Maximum average running time of 1.24 s. These results show two things: first, EPO is useful not only in raising average performance but also in raising the quality and reliability of solutions in both synthetic and pragmatic optimization applications.