Aldawsari Hamad
Department of Computer Science, Haql University College, University of Tabuk, Tabuk, Saudi Arabia.
Sci Rep. 2025 Mar 27;15(1):10522. doi: 10.1038/s41598-025-93458-8.
Various sectors and applications, including machine learning, data mining, operations research, economical problem, and science, can be structured as multi-objective optimization problems. This study introduces a novel multi-objective algorithm based on the recently developed parrot optimizer (PO) called MOPO. An external repository matrix i.e. "archive" is incorporated with the PO so that maintain the Pareto optimal solutions achieved. The MOPO utilizes the elitist non-dominated sorting, to maintain the diversity among the optimal set of solutions, further the mutate-leaders strategy is proposed to to strengthen the diversity of obtained Pareto solutions and mitigates the risk of local minima. The efficacy of the MOPO is assessed through optimizing two categories of multi-objective, include twenty benchmark test suite from the IEEE CEC'20, and real-world multi-objective design challenge, through optimizing the sensor placement in helicopter main rotor blade. The MOPO is compared against nine well-known, recent and robust multi-objective optimization algorithms. Various quantative and qualitative metrics are employed to conduct a comprehensive examination of the results; further the Friedman test and Wilcoxon test are applied on results of the four performance metrics i.e. PSP, HV, IGDf and IDGX, it demonstrates that the MOPO performed comparably to other algorithms on the most test methods, and achieved the first rank among other competitors. The Wilcoxon test exhibit the significant variance of MOPO rather competitors on p-value = 0.05. The MOPO takes average execution time less than MOSMA, SPEA2, MOPSO by 20% rate.
包括机器学习、数据挖掘、运筹学、经济问题和科学在内的各个领域和应用，都可以构建为多目标优化问题。本研究基于最近开发的鹦鹉优化器（PO）引入了一种名为MOPO的新型多目标算法。将一个外部存储库矩阵即“存档”与PO相结合，以维护所获得的帕累托最优解。MOPO利用精英非支配排序来维持最优解集之间的多样性，此外还提出了变异领导者策略，以增强所获得的帕累托解的多样性，并降低局部最小值的风险。通过优化两类多目标问题来评估MOPO的有效性，其中包括来自IEEE CEC'20的20个基准测试套件以及实际多目标设计挑战，即通过优化直升机主旋翼叶片中的传感器布局。将MOPO与九种著名的、近期的且强大的多目标优化算法进行比较。采用各种定量和定性指标对结果进行全面检验；此外，对四个性能指标即PSP、HV、IGDf和IDGX的结果应用弗里德曼检验和威尔科克森检验，结果表明MOPO在大多数测试方法上与其他算法表现相当，并在其他竞争对手中排名第一。威尔科克森检验表明，在p值 = 0.05时，MOPO与竞争对手相比存在显著差异。MOPO的平均执行时间比MOSMA、SPEA2、MOPSO少20%。
