HGHD:一种基于超图的高维空间数据聚类算法  

HGHD:An Algorithm for Clustering Data in High Dimensional Space Based on Hypergraph

作　　者：沙金[1] 张翠肖[1] 贾玉锋[1] 胡迎新[1] 

机构地区：[1]石家庄铁道学院计算机系,河北石家庄050043

出　　处：《微电子学与计算机》2006年第6期185-187,共3页Microelectronics & Computer

摘　　要：传统聚类算法无法有效地处理现实世界中存在许多高维空间数据。为此,提出一种基于超图模式的高维空间数据聚类算法HGHD,通过数据集中的数据及其间关系建立超图模型,并应用超图划分进行聚类,从而把一个求解高维空间数据聚类问题转换为一个超图分割寻优问题。该方法采用自底向上的分层思想,相对于传统方法最大的优势是不需要降维,直接用超图模式描述原始数据之间的关系,能产生高质量的聚类结果。Most of the traditional algorithms fail to produce meaningful clusters in high dimension space data sets. Therefore, a method is proposed for clustering data in high dimensional space. It maps the data and the relationship in the data into a hypergraph, cluster data by parting this hypergraph, the problem of solving the data clustering in high dimensional space is formulated as a hypergraph optimal partition problem. One of the major advantages of this scheme over traditional clustering schemes is that it does not require dimensionality reduction, It uses the hypergraph model to represent relations among the original data items. It can produce high quality cluster effectively.

关 键 词：超图模式 高维空间数据 数据聚类 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]