Mining frequent subgraphs (FSG) is one form of graph mining for which only main memory algorithms exist currently. There are many applications in social networks, biology, computer networks, chemistry and the World Wide Web that require mining of frequent subgraphs. The focus of this paper is to apply relational database techniques to support frequent subgraph mining. Some of the computations, such as duplicate elimination, canonical labeling, and isomorphism checking are not straightforward using SQL. The contribution of this paper is to efficiently map complex computations to relational operators. Unlike the main memory counter parts of FSG, our approach addresses the most general graph representation including multiple edges between any two vertices, bi-directional edges, and cycles. Experimental evaluation of the proposed approach is also presented in the paper.
展开▼