Mining Trajectory Patterns by Incorporating Temporal Properties

Spatio-temporal patterns extracted from historical trajectories of moving objects unveil important knowledge about movement behavior for high quality LBS services. Existing approaches transform trajectories into sequences of regional symbols and discover frequent subsequences by applying conventional sequential pattern mining algorithms. However, spatio-temporal correlations in the original data can be lost due to the inappropriate spatial and temporal approximation. In this paper, we propose a new algorithm for retrieving spatiotemporal patterns in trajectory data. We study the problem that inefficient description of temporal information decreases the mining efficiency and the interpretability of extracted patterns. Using line simplification, proposed method first abstracts trajectories into spatial approximations considering movements of objects, and clusters them into spatio-temporal regions incorporating temporal constraints. Finally frequent spatiotemporal patterns are extracted from discretized sequences, based on a prefix-based projection approach. We experimentally analyze that the proposed method improves mining performance and derives more intuitive patterns. Keywords-component; Spatio-Temporal data mining, Pattern mining, Trajectories