基于文化算法的模糊聚类分析

Fuzzy Clustering Analysis Based on Cultural Algorithms

分析了模糊c均值(Fuzzy C Mean)聚类算法存在的不足,提出了基于文化算法的新型聚类算法.文化算法具有双层机构的特性,能从进化种群空间中获得求解问题的知识(即信仰)来指导搜索过程,从而具有较好的全局寻优性能.仿真实验表明,基于文化算法的聚类分析方法能在一定程度上避免FCM算法对初始值敏感和容易陷入局部最优解的缺陷,具有较好的聚类结果,且需要较少的计算量.

A new clustering algorithms based on culture algorithms is proposed after analyzing the disadvantages of the Fuzzy C Mean （FCM）. Cultural algorithms has the feature of dual inheritance systems. It can obtain the knowledge （belief） of solved questions from Population Space to guide the searching process, so it has greater searching capability globally. By comparison results, a new clustering algorithms based on culture algorithms, to a certain extent, can avoid the sensitivity of initial value and the weakness of easily got struck in local optimum. Experiments show that new algorithms has highly clustering results and a relatively low compulational cost.