Querying Shortest Path Distance with Bounded Errors in Large Graphs

Miao Qiao, Hong Cheng, Jeffrey Xu Yu

Abstract: Shortest paths and shortest path distances are important primary queries for users to query in a large graph. In this paper, we propose a new approach to answer shortest path and shortest path distance queries efficiently with an error bound. The error bound is controlled by a user-specified parameter, and the online query efficiency is achieved with prepossessing offline. In the offline preprocessing, we take a reference node embedding approach which computes the single-source shortest paths from each reference node to all the other nodes. To guarantee the user-specified error bound, we design a novel coverage-based reference node selection strategy, and show that selecting the optimal set of reference nodes is NP-hard. We propose a greedy selection algorithm which exploits the submodular property of the formulated objective function, and use a graph partitioning-based heuristic to further reduce the offline computational complexity of reference node embedding.