Top- $k$ Subgraph Query Based on Frequent Structure in Large-Scale Dynamic Graphs

Frequent structures have emerged as resolving the structural pattern mining issues, such as chemistry, Web applications, and other related problems. Top-k subgraph query as an important technology of graph search is widely used in emerging fields. Different from the traditional subgraph query, the query requirement studied in this paper has two unique properties: 1) data graphs change dynamically over time, including vertices/edges insert, delete, and frequent labels update and 2) query with limits of labels and k. Existing graph index and query techniques rarely considered these. Therefore, this paper proposes a method called frequent subgraph dynamic Top-k query with label value constraint to address both the challenges. We propose a two-level index with the frequent structure mapping and label value aggregation, which locates the query structure quickly, then prunes, and filters according to the constraint to narrow search range and improve query efficiency. The method uses a two-level index to filter the initial result and combines the dynamic changes of graphs to modify query results. In addition, it also proposed an incremental dynamic maintenance strategy over the proposed index, which only updated the partial contents to avoid the high cost caused by global update. The experimental results demonstrate that the proposed Top-k query method outperforms the baseline approaches up to 37%.