Trajectory similarity measurement: An enhanced maximal travel match method
Fahime Karami, Mohammad Reza Malek · Transactions in GIS · 2021
Abstract Trajectory similarity measurement is a vital and widely used step in many applications, including recommendation systems. In the discipline of similarity measurement, research has mostly been focused on raw trajectories, consisting of location and time‐stamp information. Due to the explosion in the use of the internet and location‐based social networks, raw trajectories can be easily enriched with semantic information. Nevertheless, few attempts have been made to apply semantic and location information during similarity measurement. In light of this, we present a new similarity measure called the enhanced maximal travel match (EMTM) in this article. The proposed EMTM improves on conventional maximal travel match (MTM) methods by simultaneously considering the location and place category as the most basic semantic information. EMTM first generates a new place category‐location hierarchical framework (CLHF) for each trajectory. Subsequently, it identifies MTMs at each layer of hierarchy to explore the similarity between each pair of CLHFs. Finally, the proposed method calculates similarity scores based on the identified MTMs. In experiments on a Foursquare data set, EMTM outperformed the conventional MTM methods by more than 50% in terms of mean average precision. Additionally, EMTM was the most accurate technique among four state‐of‐the‐art methods.