Fast subtrajectory similarity search in road networks under weighted edit distance constraints

Abstract

In this paper, we address a similarity search problem for spatial trajectories in road networks. In particular, we focus on the subtrajectory similarity search problem, which involves finding in a database the subtrajectories similar to a query trajectory. A key feature of our approach is that we do not focus on a specific similarity function; instead, we consider weighted edit distance (WED), a class of similarity functions which allows user-defined cost functions and hence includes several important similarity functions such as EDR and ERP. We model trajectories as strings, and propose a generic solution which is able to deal with any similarity function belonging to the class of WED. By employing the filter-and-verify strategy, we introduce subsequence filtering to efficiently prunes trajectories and find candidates. In order to choose a proper subsequence to optimize the candidate number, we model the choice as a discrete optimization problem (NP-hard) and compute it using a 2-approximation algorithm. To verify candidates, we design bidirectional tries , with which the verification starts from promising positions and leverage the shared segments of trajectories and the sparsity of road networks for speed-up. Experiments are conducted on large datasets to demonstrate the effectiveness of WED and the efficiency of our method for various similarity functions under WED.