CAJ | 학술논문

聚类集成算法通常对聚类成员差异性要求较高,导致算法在生成聚类成员阶段计算复杂度提高.针对该问题提出了一种基于遗传算法的聚类集成方法CEGA,不考虑聚类成员的差异性,而是利用目标函数将聚类问题转化为聚类成员的优化问题,充分利用遗传算法内在的并行性和全局寻优能力,对聚类成员进行优化组合,并以得到的最优染色体作为聚类集成最终结果.分析了CEGA的复杂度及适用范围,并利用UCI数据库中部分数据集进行实验,实验结果表明这种聚类集成方法的有效性.
취류집성산법통상대취류성원차이성요구교고,도치산법재생성취류성원계단계산복잡도제고.침대해문제제출료일충기우유전산법적취류집성방법CEGA,불고필취류성원적차이성,이시이용목표함수장취류문제전화위취류성원적우화문제,충분이용유전산법내재적병행성화전국심우능력,대취류성원진행우화조합,병이득도적최우염색체작위취류집성최종결과.분석료CEGA적복잡도급괄용범위,병이용UCI수거고중부분수거집진행실험,실험결과표명저충취류집성방법적유효성.
Clustering ensemble algorithms require higher differences among clustering components, which induce higher com-plexity during the generating phase of clustering components. This paper proposes a new model for Clustering Ensemble based on Genetic Algorithm(CEGA), which does not need to consider the differences between clustering components, but translates clustering into optimization of clustering components by calculating target function, and optimizes the grouping of clustering components by genetic algorithms. CEGA sets the final optimal chromosome to be the result of clustering and its complexity and application are also analyzed. Experimental results demonstrate the effectiveness of the proposed method on several UCI datasets.