一种改进的图聚类的相异度度量方法

AN IMPROVED DISSIMILARITY METRICS APPROACH FOR GRAPH CLUSTERING

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.