Mining closed frequent free trees in graph databases

Abstract

Free tree, as a special graph which is connected, undirected and acyclic, has been extensively used in bioinformatics, pattern recognition, computer networks, XML databases, etc. Recent research on structural pattern mining has focused on an important problem of discovering frequent free trees in large graph databases. However, it can be prohibitive due to the presence of an exponential number of frequent free trees in the graph database. In this paper, we propose a computationally efficient algorithm that discovers only closed frequent free trees in a database of labeled graphs. A free tree t is closed if there exist no supertrees of t that has the same frequency of t. Two pruning algorithms, the safe position pruning and the safe label pruning, are proposed to efficiently detect unsatisfactory search spaces with no closed frequent free trees generated. Based on the special characteristics of free tree, the automorphism-based pruning and the canonical mapping-based pruning are introduced to facilitate the mining process. Our performance study shows that our algorithm not only reduces the number of false positives generated but also improves the mining efficiency, especially in the presence of large frequent free tree patterns in the graph database. © Springer-Verlag Berlin Heidelberg 2007.