摘要
本文提出了一种新的目标函数,通过此函数的极小化,得到了模糊聚类的一种方法.此方法可对通常的模糊ISODATA方法进行改进,不需要计算每类的聚类中心,而直接通过迭代求得最优的分类矩阵,因此计算量比模糊ISODATA要少.此外,对模糊ISODATA算法给出了一种修改意见.
In this paper,a new kind of objective function is established. By minimizing the objective function, we obtain a fuzzy clustering algorithm, by means of which we can get the optimal classification matrix without calculating the cluster center. Therefore, the calculation steps of the method are less than those of the fuzzy ISODATA. In addition, some improvement suggestions about fuzzy ISODATA are given.
关键词
目标函数
模糊集
聚类中心
Objective function
Fuzzy clustering analysis
Cluster center
Fyzzy ISODATA
