User oriented trajectory similarity search

Trajectory similarity search studies the problem of finding a trajectory from the database such the found trajectory most similar to the query trajectory. Past research mainly focused on two aspects: shape similarity search and semantic similarity search, leaving personalized similarity search untouched. In this paper, we propose a new query which takes user's preference into consideration to provide personalized searching. We define a new data model for this query and identify the efficiency issue as the key challenge: given a user specified trajectory, how to efficiently retrieve the most similar trajectory from the database. By taking advantage of the spatial localities, we develop a two-phase algorithm to tame this challenge. Two optimized strategies are also developed to speed up the query process. Both the theoretical analysis and the experiments demonstrate the high efficiency of the proposed method.