基于相对密度的增量式聚类算法  

Relative Density Based Incremental Clustering Algorithm

作　　者：刘青宝[1] 侯东风[1] 邓苏[1] 张维明[1] 

机构地区：[1]国防科技大学信息系统与管理学院,湖南长沙410073

出　　处：《国防科技大学学报》2006年第5期73-79,共7页Journal of National University of Defense Technology

基　　金：国家自然科学基金资助项目(60172012)

摘　　要：基于聚类的相对性原则:簇内对象具有较高的相似度,而簇间对象则相反,提出一种基于相对密度的增量式聚类算法,它继承了基于绝对密度聚类算法的抗噪声能力强、能发现任意形状簇等优点[1],并有效解决了聚类结果对参数设置过于敏感、参数值难以确定以及高密度簇完全被相连的低密度簇所包含等问题。同时,通过定义新增对象的影响集和种子集能够有效支持增量式聚类。A new incremental clustering algorithm is proposed in this paper based on the relativity principle, which means that the similarities of objects in the same cluster is higher than those among different clusters. This approach not only inherits the advantages of absolute density based algorithms which can discover arbitrary shape clusters and are insensitive to noises , but also efficiently solves the following common problems： clustering results are very sensitive to the user-deflned parameters, reasonable parameters are hard to be determined, and high density clusters are contained fully in coterminous low density clusters. With this approach, incremental clustering can also be supported effectively by defining the affected sets and seed sets of the updating objects in this approach.

关 键 词：增量式聚类 K近邻 聚类参数 相对密度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]