A general powerful graph pattern matching system for data analysis

Abstract

Graph pattern matching is a powerful mechanism for searching on network data. Most of the graph pattern matching tools available are based on subgraph isomorphism, i.e. finding a one-to-one correspondence between nodes of a query graph and nodes of a target graph. Often this approach is not flexible enough, since it does not allow the query to represent sets of nodes of any size that share the same profile of connectivity. We propose a novel powerful graph matching approach that overcomes the existing limits and allows the user to define complex patterns in a simple and intuitive way. In our approach, queries are expressed as graphs, whose nodes and edges specify structural constraints and filtering criteria. We show that, despite its simplicity, the proposed approach can solve a large variety of practical problems.