K hops frequent subgraphs mining for large attribute graph

Abstract

Attribute Graphs are widely used to describe complex data in many applications such as bio-informatics and social network. With the rapid growth of scale for graph data, traditional solutions for mining frequent subgraphs cannot performed well in large attribute graph because of time-consuming candidates generation and isomorphism testing. In this paper, we investigate the problem for k hops subgraph mining in large attribute graph. The attribute graph is transformed into labeled graphs by projection for each attribute. K hops frequent subgraph mining algorithm FSGen consists of three procedures is performed. Firstly, frequent vertices and edges will be extended to frequent subgraphs from root vertices. Secondly, frequent edges joining frequent vertices will be added into extended subgraphs. Thirdly, if necessary, isomorphism testing will be used to summarize frequent subgraphs based on Graph Edit Distance. Then, frequent labeled subgraphs will be merged into attribute subgraphs by integration according to designated attributes. The complexity of our mechanism is approximately O(2n ), which is more efficient than existing algorithms. Real data sets are applied in experiments to demonstrate the efficiency and effectiveness of our technique. © 2013 Springer-Verlag.