基于样本最大分类信息的聚类有效性函数

A Cluster Validity Function Based on Maximum Classification Information

根据每一样本的最大分类信息 ,对 Bezdek定义的划分系数进行修改 ,提出了一个新的聚类有效性函数 ,讨论了该函数的若干性质。实验结果表明在某些情形该函数的判决效果优于划分系数。

Based on the maximum classification information, a new cluster validity function, which modified Bezdek's partition coefficient, is defined. Some properties are studied and comparison experiments shown that this function is superior to partition coefficient in some cases.