一种改进的图聚类的相异度度量方法  

AN IMPROVED DISSIMILARITY METRICS APPROACH FOR GRAPH CLUSTERING

作　　者：王小黎[1] 

机构地区：[1]中原工学院经济管理学院,河南郑州450007

出　　处：《计算机应用与软件》2011年第5期139-141,共3页Computer Applications and Software

基　　金：河南省自然科学基金(0411011400)

摘　　要：采用凝聚聚类思想进行聚类分析,通过对距离最小的两个类进行循环合并,从而改进了最大最小关联度图聚类方法在针对图中两个邻接结点的度都非常大,而它们连接相同结点的数目极少的问题时,普通聚类方法难以解决两个结点的关联度很大,而实际上相似性并不高的情况,最后通过实例检验了凝聚聚类算法的正确性和有效性。In this paper the agglomerative clustering concept is employed to conduct the clustering analysis,by combining in circulation two categories with least distance,the problem of the max-min correlativity graph clustering method is improved,of which when aiming at the situation that in the graph two neighbouring nodes both have quite big correlativity but the same nodes they connecting to are extremely few,then usual clustering method is hard to solve it that the two nodes are at big correlativity whereas actually their similarities are low.At last,the correctness and effectiveness of the agglomerative clustering algorithm is validated with experiment.

关 键 词：相异度 度量 层次聚类方法 模块性 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]