基于扩展凝聚点和网格的增量聚类算法  

An incremental clustering algorithm based on extended condensation point and grid

作　　者：陈卓[1] 贺明霞[1] 刘相双[2] 

机构地区：[1]中国海洋大学海洋遥感所海洋遥感教育部重点实验室,山东青岛266071 [2]中国海洋大学计算机科学系,山东青岛266071

出　　处：《哈尔滨工业大学学报》2006年第8期1382-1385,1398,共5页Journal of Harbin Institute of Technology

基　　金：国家自然科学基金资助项目(60374031)

摘　　要：提出一种基于扩展凝聚点和网格的快速聚类算法CECPG(c lustering using extended condensation pointand grid).在CECPG算法的基础上提出一种基于扩展凝聚点和网格的增量聚类算法ICECPG(increm entalc lustering using extended condensation point and grid).通过扩张凝聚点准确反映数据空间的几何特征,然后采用网格和密度相结合的方法,利用爬山法和连通性原理进行聚类处理,并在差分数据的指导下进行增量聚类.实验结果证明,CECPG算法的聚类效果优于模糊聚类算法FCM和C lique算法.A new kind of clustering algorithm called CECPG （clustering using extended condensation point and grid） is presented. Based on CECPG, an incremental clustering algorithm called ICECPG （incremental clustering using extended condensation point and grid） is presented in order to realize the real time clustering of the dynamic data. The creativities of these algorithms are the capturing of the shape of data space by extended condensation points, and then the using of grid - based and density - based clustering methods based on the theories of climbing hill algorithm and connectedness to cluster the data, guided by the difference data to implement incremental cluster. Experimental results confirm that the executing efficiency of CECPG is much better than Fuzzy c - Means algorithm and Clique algorithm.

关 键 词：聚类 增量聚类 扩展凝聚点 爬山法 网格 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]