新的聚类方法综合评价血脂指标

本文结合中心聚类和模糊聚类提出了一个新的聚类方法。该法避免了在迭代过程中出现局部最小问题,并可给出每个样本归入每一类的成分矩阵,因而提供更多的聚类程度的信息。应用这种方法对正常人和冠心病患者的血脂指标进行了综合评价,结果满意。

New Cluster Analysis Applied to Comprehesive Evajuation Of Blood-Lipoid parameters

In this paper, a new method is proposed, combining fuzzy cluster with center cluster. The advantage of this method is that it produces the membership matrix that measures the degree of membership of each objects to each of the subgroups and thus gives useful imformation about how natural the cluster of objects into subgroups is. Another advantage is it can avoid local minima in an iterative process. A data matrix of blood parameters from patients of the coronary heart diseases and normal persons is evaluated by using this method with satisfactory results comparing with discriminant analysis.