LCGMiner: levelwise closed graph pattern mining from large databases

Abstract

LCGMiner (levelwise closed graph pattern miner) is proposed to improve CloseGraph (Yan and Han, 2003) in discovering frequent closed sub graphs. Frequent closed edgesets with the same extended vertexsets are expanded in pattern generation compared to one edge or one vertex in traditional methods. Experiments on synthetic datasets as well as a real NIH dataset demonstrates that our algorithm outperforms CloseGraph and gSpan.