利用划分方法进行混合数据聚类

目前常用的几种基于划分的聚类方法主要处理数值型数据,能有效处理实际应用领域中常用的包括数值和符号混合数据的聚类算法则较少。基于此问题,文章根据k均值、k中心点和k众数等基于划分的聚类方法各自的特点,对其进行集成与改进,提出一种能够应用于混合类型数据的聚类分析方法,即将所有的混合类型变量转换到共同的标度区间0.0,1....

Hybrid Category Data Clustering through Partitioning Methods

The usual clustering methods based on partitioning mainly process numerical data and it is lack of the clustering method that can deal with hybrid category data. Because of these problems, this paper integrates and improves the traditional and classical clustering methods those are k-means, k-medoids and k-modes in order to propose a method that can solve the cluster analysis about hybrid category data according to those traditional methods’ characteristics. This paper’s method converts all hybrid category data to same scale range between 0.0 to1.0 in order to computes the dissimilarity according to the compositive formula and updates each kind data of clustering centers independently.