Mining Temporal Association Rules of an Attributed Graph Sequence at High Conceptual Levels

To find more informative knowledge from a transactional database, some methods for mining association rules at different conceptual hierarchy have been proposed. For a given attributed graph sequence, how to mine temporal association rules at high conceptual level is the task of this paper, which has not been ever studied. In order to formalize the conceptual hierarchy of an attributed graph sequence, we do the clustering of the vertexes of each attributed graph in the given attributed graph sequence by the k-means algorithm. Viewing each cluster as a vertex, a non-attributed graph sequence can be formed. A new mining algorithm designed based on the Apriori algorithm is able to mining the temporal association rules at high conceptual level for the given attributed graph sequence. Experiments presented in this paper show the good performance of the proposed algorithm.