节点文献
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
【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.
【Abstract】 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.
【Key words】 Clustering; Minimal Spanning Tree; Linear Constraint; Spatial Data Mining;
- 【文献出处】 模式识别与人工智能 ,Pattern Recognition and Artificial Intelligence , 编辑部邮箱 ,2002年04期
- 【分类号】TP311.13
- 【被引频次】3
- 【下载频次】115