关键词: 图挖掘, 中心问题, 相似模式, 关联相似模式

Abstract: Many of the previous studies show convincing arguments that mining frequent subgraphs is especially useful.Many hidden frequent patterns which are very interesting can not be found by mining single graph.Therefore，it needs mine frequent patterns from multiple graphs.Previous studies as quasi-clique have little success with the hub problem.This paper introduces a new conception correlated-quasi-clique and develops a novel algorithm，CoClique，to address the hub problem and improve the efficiency of frequent correlated-quasi-cliques mining.Meanwhile，it exploits several effective techniques to prune the search space.An extensive experimental evaluation on real databases demonstrates that the algorithm outperforms previous methods.

Key words: graph mining, hub problem, quasi-clique, correlated-quasi-clique