Approximate Subgraph Matching Query over Large Graph

Approximate subgraph matching query is increasingly adopted to retrieve labeled, heterogeneous networks with millions of vertices and edges. Those networks are usually noisy and lack of fixed schema. Previous exact subgraph matching query (such as subgraph isomorphism) and approximate matching aimed at small proprietary network are not practicable. Recently approximate subgraph matching over large graph usually reduces match accuracy to ensure query efficiency. In this paper, We present a novel approximate subgraph matching query method. We propose a similarity score function to measure the subgraph match quality. Based on it, we adopt a two-step-strategy in subgraph matching query processing: candidate selection and query processing. And we employ an indexing technique to improve query efficiency. We experimentally evaluate our method on query efficiency and effectiveness. The results demonstrate that our method outperforms state-of-the-art method NeMa especially on efficiency.