基于特征向量的分布式聚类算法  

Distributed clustering algorithm based on feature vector

作　　者：李锁花[1] 孙志挥[1] 周晓云[1] 

机构地区：[1]东南大学计算机科学与工程系,江苏南京210096

出　　处：《计算机应用》2006年第2期379-382,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(70371015)

摘　　要：提出了一种新的表达数据集的方法———特征向量,它通过坐标和密度描述了某一密集空间,以较少的数据量反映站点数据的分布特性。在此基础上提出了一种基于特征向量的分布式聚类算法———DCBFV(D istributed C lustering Based on Feature Vector),该算法可有效降低网络通信量,能够对任意形状分布的数据进行聚类,提高了分布式聚类的时空效率和性能。理论分析和实验结果表明DCBFV是高效可行的。Distributed clustering is a new research field of data mining. A new idea to sketch site data called feature vector was proposed, which described a dense space through coordinate and density, so captured distribution characteristic of data efficiently. Then a novel approach named DCBFV （ distributed clustering based on feature vector） based on feature vector was proposed. DCBFV can decrease network overload, discover clusters with arbitrary shape, and improve the quality of global clustering effectively. Both theory analysis and experimental results confirm that DCBFV is feasible and effective.

关 键 词：数据挖掘 分布式聚类 特征向量 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]