A New Approach to Isomorphism in Attributed Graphs

Attributed graphs are widely used in many application domains, for example to model social networks. An attributed graph is a graph in which vertices and edges may have types and other attributes. Different query models have been developed to obtain information from attributed graphs. One of the most important is graph pattern matching, which is the problem of finding all the instances of the pattern graph P in the attributed graph G under graph isomorphism. A pattern graph may specify both structural requirements and predicates on attributes of the graph elements. We propose a novel technique that linearizes the pattern graph and matches such linearization against the attributed graph. We derive heuristics to produce a linearization that places selective predicates at the beginning. We implement the algorithm and our results show that our optimizations based on the attributed graph statistics are effective in querying attributed graphs.