Title:Compact Neighborhood Index for Subgraph Queries in Massive Graphs

Abstract:Subgraph queries also known as subgraph isomorphism search is a fundamental problem in querying graph-like structured data. It consists to enumerate the subgraphs of a data graph that match a query graph. This problem arises in many real-world applications related to query processing or pattern recognition such as computer vision, social network analysis, bioinformatic and big data analytic. Subgraph isomorphism search knows a lot of investigations and solutions mainly because of its importance and use but also because of its NP-completeness. Existing solutions use filtering mechanisms and optimise the order within witch the query vertices are matched on the data vertices to obtain acceptable processing times. However, existing approaches are iterative and generate several intermediate results. They also require that the data graph is loaded in main memory and consequently are not adapted to large graphs that do not fit into memory or are accessed by streams. To tackle this problem, we propose a new approach based on concepts widely different from existing works. Our approach distills the semantic and topological information that surround a vertex into a simple integer. This simple vertex encoding that can be computed and updated incrementally reduces considerably intermediate results and avoid to load the entire data graph into main memory. We evaluate our approach on several real-word datasets. The experimental results show that our approach is efficient and scalable.