Discovery of Travelling Companions from Trajectories with Different Sampling Rates

Discovery of travelling companions from trajectories can provide empirical support for various applications, such as COVID-19 contact tracing, suspects tracking and detection, tourist behavior analysis, etc. One challenge is trajectories of travelling companions are from different data sets with different sampling rates and granularities. Most current researches for discovery of travelling companions focus on using snapshot-based clustering methods to identify travelling groups, or using trajectory similarity algorithms to mine companion relationships. However, the constantly changing sampling rate limits the application of clustering methods in the companion relationship mining. Although some similarity algorithms can mitigate this negative impact, they usually focus on the spatial distribution of trajectories and the time complexity is very high. In this paper, we designed a Spatio-Temporal Trajectory Companion DEtection Framework (STCDEF) to detect travelling companions from trajectories with different sampling rates, which can effectively reduce the time consumption caused by the matching mechanism. Within the STCDEF, an approximate trajectory similarity algorithm, Fast Spatio-Temporal Similarity (FSTS) measure, is presented. Moreover, the concept of Mutual Following Degree (MFD) is introduced into STCDEF to detect travelling companions with FSTS, so as to further improve the efficiency when dealing with trajectories of varying sampling rates.