Towards Recurring Co-traveling Pattern Detection: A Summary of Results

Given trajectories or location traces and user-specified thresholds, we investigate algorithms to detect recurring co-traveling patterns. For example, a school bus transports students between a neighborhood and a school. The problem is important for its societal applications in anomaly detection, synthetic trajectory and location trace data evaluation, and transportation planning. For example, deviations from recurring co-traveling groups naturally highlight anomalies, such as unexpected disruptions in commuting flows or rare co-traveling events. The problem is challenging due to the need to model recurring co-traveling routes and process an exponentially large number of candidate groups. Existing spatiotemporal data mining methods primarily focus on detecting co-occurrence relationships, but do not identify recurring co-traveling routes with specific travel areas. To overcome these limitations, we propose a novel recurring co-traveling group interest measure and Recurring Co-traveling Pattern Detection (RCPD) algorithms. We employ a divide-and-conquer method and spatial indices to improve computation efficiency. We also provide theoretical proofs that the proposed interest measure has the anti-monotone property, allowing early pruning, and the proposed algorithm is correct and complete. We evaluate our methods using real and synthetic trajectory/location trace data, as well as a case study on anomaly detection.