一种基于划分的层次聚类算法  

Hierarchical clustering algorithm based on partition.

作　　者：沈洁[1] 赵雷[1] 杨季文[1] 李榕[1] 

机构地区：[1]苏州大学计算机科学与技术学院,江苏苏州215006

出　　处：《计算机工程与应用》2007年第31期175-177,共3页Computer Engineering and Applications

基　　金：江苏省计算机信息处理技术重点实验室开放基金(No.JSK0604)

摘　　要：CURE算法是针对大规模数据聚类算法的典型代表。提出了一种新的算法K-CURE,该方法基于划分思想对CURE算法作了改进,同时给出了在聚类中剔除孤立点的时机选择方法。测试表明,改进后的算法效率明显高于原算法,且聚类效果良好。CURE is a typical clustering algorithm that is introduced in this article to improve the CURE based on occasion of eliminating outlier during clustering.Experiments designed for the mining of mass data.A new algorithm K-CURE is partition.A method is also described to explain how to choose the indicate that the improved algorithm does improve the CURE in both efficiency and effectiveness.

关 键 词：数据挖掘 层次聚类 代表对象 CURE 孤立点 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]