Time-Efficient Discovery of Moving Object Groups from Trajectory Data

Abstract

The advent of numerous mobile devices and location acquiring technologies like GPS give rise to a massive amount of spatio-temporal data. These devices leave the traces of their positions in the form of a trajectory, which imparts valuable information regarding an object’s mobility, as it moves with time. The identification of object groups has copious applications in various domains, like transport system, event prediction, scientific studies, etc. All the state-of-the-art algorithms for discovering object groups use DBSCAN Lo et al. (Kdd 96:226–231, 1996) [1] for clustering spatio-temporal data. However, the time cost for DBSCAN is $O(n^{2})$ which can be futile for streaming data. Our work lies in improving the time complexity of the buddy-based traveling companion (a certain type of moving object group) discovery algorithm Tang et al. (ICDE, 2012) [2], Tang et al. (ACM Transactions on Intelligent Systems and Technology (TIST) 5(1):3, 2003) [3] by incorporating the grid based clustering algorithm Gunawan (PhD thesis, Masters thesis, Technische University Eindhoven, 2013) [4], which takes O(nlogn) time. We also establish a novel concept of varying density with increasing snapshots.