基于模糊C均值的空间聚类方法研究  

Spatial clustering method based on fuzzy C-mean

作　　者：王海军[1] 邓羽[2,3] 刘佳[1] 

机构地区：[1]武汉大学资源与环境科学学院,武汉430079 [2]中国科学院地理科学与资源研究所,北京100101 [3]中国科学院研究生院,北京100049

出　　处：《华中师范大学学报（自然科学版）》2009年第1期152-155,共4页Journal of Central China Normal University：Natural Sciences

基　　金：国家自然科学基金资助项目(40871179);中国科学院科技创新与社会实践资助专项

摘　　要：空间聚类是空间数据挖掘和知识发现的的主要方法之一."基于模糊C均值的空间聚类方法"可以广泛地应用到对空间数据挖掘和知识发现中的分类分级研究.该方法的基本思想是:首先,设定聚类数目和模糊度常数,并初始化各个聚类中心,每一个数据点按照一定的模糊隶属度隶属于某一聚类中心;然后,逐步进行循环迭代,改变目标函数值以及各个隶属度,并决定新一级聚类中心;最后,当目标函数收敛或者数据点的隶属度保持恒定时,就得到了输入数据的最终聚类中心,从而完成了模糊聚类划分.该文给出了该方法的实现步骤,并以实例验证了方法的可行性和科学性,取得了很好的效果.Spatial clustering is the main methods of Spatial Data Mining and Knowledge Discovery. The purpose of this paper is to set forth Spatial Clustering Method based on Fuzzy Cmean, which can be widely applied to the research on classification and hierarchy in realm of Spatial Data Mining and Knowledge Discovery. The basic idea of this method is： First, set the number of clustering center and fuzziness index, initialize clustering center and make each data points belong to a particular cluster center according to certain fuzzy membership. Then, bring iterative cycle gradually, change in objective function value as well as membership and decide the new clustering center. Finally, when the objective function or data points of convergence membership maintains a constant; we can receive the final clustering center, thus completing the fuzzy clustering. This paper presents implementation steps of the method, verifies the feasible and scientific significance by examples, and acquires very good results.

关 键 词：空间聚类 模糊C均值 空间数据挖掘 知识发现 

分 类 号：P208[天文地球—地图制图学与地理信息工程]