一种高维空间数据的模糊聚类算法  

A fuzzy clustering algorithm of high dimensional spatial data

作　　者：杨悦[1] 张健沛[1] 李忠伟[1] 

机构地区：[1]哈尔滨工程大学计算机科学与技术学院,黑龙江哈尔滨150001

出　　处：《哈尔滨工程大学学报》2006年第B07期485-488,共4页Journal of Harbin Engineering University

基　　金：黑龙江省自然科学基金资助项目（F2005-02）.

摘　　要：针对传统的基于网格-密度的空间聚类方法容易产生不平滑聚类、非坐标轴方向过度聚类以及聚类边界判断模糊的问题，本文提出了一种高维空间数据的模糊聚类算法．该算法通过扩展网格区域，用模糊集的隶属度对基本区域及模糊扩展区域内的数据点进行计数，考虑了相邻网格对当前考察网格内数据点的影响，避免了不平滑聚类想象：同时，通过对相邻网格重新定义扩展了聚类算法的执行方向，有效缓解了过度聚类以及聚类边界模糊的问题．实验结果表明，该方法克服了传统聚类方法的不足，空间高维数据聚类结果的质量得到了改善．In order to solve the problems of traditional grids and density based algorithm, which contain un-smoothness clustering, unclear boundary of spatial clustering results and overly clustering at un-coordinates axes directions, a fuzzy algorithm of high dimensional spatial clustering is proposed in this paper. This algorithm extends the region of the grids, uses membership degrees to count the data points in basic regions and fuzzy extended regions and uses density threshold for clustering, considers the affect of spatial data in adjacent grids and current grid. The clustering becomes smoother and the boundary of results becomes clearer. At the same time, the adjacent grids are re-defined in new algorithm to extend the directions of clustering operating, and the excessive small clustering is avoided efficiently on un-coordinates axes directions. The experiment results show that this new algorithm conquers the disadvantage of traditional method efficiently, and improves the quality of clustering results of high dimensional spatial data.

关 键 词：高维空间聚类 模糊集 隶属度 模糊扩展区域 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]