CAJ | 학술논문

采用精英策略的非支配排序遗传算法(NSGA-II)种群收敛分布不均匀，全局搜索能力较弱。针对该问题，基于现有的算法，提出一种基于聚类学习机制的多目标进化算法KMCNSGA-II。利用K均值聚类对目标函数和个体分别进行聚类，对聚类后的个体进行局部学习，以提高适应度。将该算法应用于经典的多目标约束和非约束测试函数中，通过收敛性指标世代距离和多样性指标?进行性能评价。实验结果表明，与NSGA-II算法相比，该算法在算法收敛性和种群多样性保持方面均有明显提高。
채용정영책략적비지배배서유전산법(NSGA-II)충군수렴분포불균균，전국수색능력교약。침대해문제，기우현유적산법，제출일충기우취류학습궤제적다목표진화산법KMCNSGA-II。이용K균치취류대목표함수화개체분별진행취류，대취류후적개체진행국부학습，이제고괄응도。장해산법응용우경전적다목표약속화비약속측시함수중，통과수렴성지표세대거리화다양성지표?진행성능평개。실험결과표명，여NSGA-II산법상비，해산법재산법수렴성화충군다양성보지방면균유명현제고。
According to the uneven distribution of population convergence and poor performance in global search of Non-dominated Sorting Genetic Algorithm II(NSGA-II), a multi-objective evolutionary algorithm, called K-means clustering non-dominated sorting genetic algorithm II(KMCNSGAII) is proposed with combining the theory and the existing algorithm. The KMCNSGAII uses K-means clustering technology and at the same time clusters both all the objective functions and individuals respectively. Then the learning and improvement method is used with respect to individuals after clustering. The KMCNSGAII algorithm is applied to several classical unconstrained and constrained test functions. Experimental results demonstrate that the KMCNSGAII achieves good results with performance evaluation about convergence indicator and diversity indicator, in convergence and diversity of population both are improved significantly compared with NSGA-II.