Extracting and Clustering Method of Web Bipartite Cores

Abstract

The paper focuses on some key problems in Web communities' discovery. Based on topic-oriented communities discovery, we analyze some insufficiencies of CBG(complete bipartite graph) in trawling method. The conception of x-coreset is introduced, instead of CBG, it is more reasonable as a signature of core of community. We construct a bipartite graph from a node x and then (i, j)pruning the graph to obtain x-coresset. By scanning topic subgraph, we can extract a set of x-coressets. Finally, a hierarchal clustering algorithm is applied to these x-cores-sets and the dendrogram of community is formed. We proved that x-cores-set, consisted of x-cores, can be calculated by a bipartite graph collected from x and (i, j)pruning. The experiment is set up on the dataset that is same as that in HITS method, except for returned pages are integrated from 4 search engines. The result shows that our algorithm is effective and efficient.