Discovery of correlated sequential subgraphs from a sequence of graphs

Abstract

Dynamic graphs or a sequence of graphs attract much attention recently. In this paper, as a first step towards finding significant patterns hidden in dynamic graphs, we consider the problem of mining successive sequence of subgraphs which appear frequently in a long sequence of graphs. In addition, to exclude insignificant patterns, we take into account the mutual dependency measured by θ-correlation coefficient among the components in patterns. An algorithm named CorSSS, which utilizes the generality ordering of patterns effectively, is developed for enumerating all frequent and correlated patterns. The effectiveness of CorSSS, is confirmed through the experiments using real datasets. © 2009 Springer.