数据聚类的共同进化方法

The Cooperative Co-evolutionary Algorithm for the Data Clustering

共同进化算法是一种新的进化算法,由于它采用了解空间分离编码,能有效地克服一般进化算法中固有的早熟收敛问题。该文针对数据聚类问题——当前数据挖掘与探查性数据分析中的一个重要课题——将数据聚类问题抽象成为一个赋值图的分割问题,应用共同进化算法来加以解决,使得聚类的结果不必依赖于初始聚类中心,并对该算法的性能加以分析。将该算法与一般的遗传算法相比较,通过实验证明了该算法的优越性能。

The cooperative co-evolutionary algorithm,which is an evolutionary algorithm that can overcome the prema-ture convergence problem,is discussed in this paper.We have studied the application of the cooperative co-evolutionary algorithm to the data-clustering problem that is an important study in the field of data mining,abstracting the problem to a labeled graph to make it easy.The capability of the algorithm is analyzed and the fine capability in improving pre-mature convergence and convergent efficiency contrast to the simple genetic algorithm are verified by simulative experi-ment at the end of paper.