一种带线性约束的最小生成树聚类方法  

A MST BASED CLUSTERING METHOD WITH LINEAR CONSTRAINT

作　　者：汪闽[1] 周成虎[1] 裴韬[1] 骆剑承[1] 

机构地区：[1]中国科学院资源与环境信息系统国家重点实验室,北京100101

出　　处：《模式识别与人工智能》2002年第4期494-497,共4页Pattern Recognition and Artificial Intelligence

基　　金：中国科学院知识创新项目(CX10G-D00-06);(KZCX1-Y-02);国家自然科学基金(40101021)

摘　　要：提出了一种带线性约束的最小生成树聚类方法,目标是聚合在空间上大致呈线性密集分布的对象.方法的基本过程是,用线性率阈值约束最小生成树打断边的选取,尽可能使每次打断都能割取出一个满足线性率大于该阈值的子树(类),直至所有合适子树都被割取,残余子树则被抛弃.对自构建数据和真实世界中地震数据的聚类实验证明了该方法的有效性和实用性.In this paper, a MST based clustering method with linear constraint is proposed, whose aim is to clustering the objects distributing densely and linearly in space. The algorithm restricts the selection of the splitting edges of the MST with linear threshold, which tries to cut off one sub-tree whose linear rate exceeds the threshold every splitting. The algorithm will stop when all the suitable sub-trees are cut off, and the remaining sub-trees are discarded. The effectiveness and practicality of our methods are validated by clustering the constructed data and the earthquake data.

关 键 词：线性约束 最小生成树 聚类方法 空间数据挖掘 空间数据库 知识发现 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]