TrajGEOS: Trajectory Graph Enhanced Orientation-based Sequential Network for Mobility Prediction

TL;DR

TrajGEOS addresses next-location prediction by integrating a global trajectory graph with hierarchical graph convolutions to learn location and long-term user embeddings, and by introducing an orientation-based mid-term preference module that leverages recent history. It fuses multi-preference signals (long-term, mid-term, short-term) in a multi-task framework that also predicts next location and next category, improving predictive accuracy over state-of-the-art baselines on three real-world LBSN datasets. The method demonstrates the value of combining graph-based relational modeling with sequence-aware orientation to capture complex mobility patterns and context, with ablations confirming the necessity of each component. The approach has practical implications for urban mobility services and location-based applications, and the authors discuss scalability and potential incorporation of social networks in future work.

Abstract

Human mobility studies how people move to access their needed resources and plays a significant role in urban planning and location-based services. As a paramount task of human mobility modeling, next location prediction is challenging because of the diversity of users' historical trajectories that gives rise to complex mobility patterns and various contexts. Deep sequential models have been widely used to predict the next location by leveraging the inherent sequentiality of trajectory data. However, they do not fully leverage the relationship between locations and fail to capture users' multi-level preferences. This work constructs a trajectory graph from users' historical traces and proposes a \textbf{Traj}ectory \textbf{G}raph \textbf{E}nhanced \textbf{O}rientation-based \textbf{S}equential network (TrajGEOS) for next-location prediction tasks. TrajGEOS introduces hierarchical graph convolution to capture location and user embeddings. Such embeddings consider not only the contextual feature of locations but also the relation between them, and serve as additional features in downstream modules. In addition, we design an orientation-based module to learn users' mid-term preferences from sequential modeling modules and their recent trajectories. Extensive experiments on three real-world LBSN datasets corroborate the value of graph and orientation-based modules and demonstrate that TrajGEOS outperforms the state-of-the-art methods on the next location prediction task.

Figures (6)

• Figure 1: (a) Implicit relationship between locations. (b) Sample users' trajectory graphs built from their historical check-in data.
• Figure 2: The architecture of TrajGEOS. It contains three modules: graph learning module, trajectory embedding module, and prediction module.
• Figure 3: Hierarchical graph convolution in the graph modeling module.
• Figure 4: (a) The cumulative distribution function of distance error($\hat{l}^{T+1}, l^{T+1}$). (b) The relationship between users' record num and the predicted accuracies of their next locations.
• Figure 5: (a) Relationship of historical location complexity and predicted accuracy. (b) Relationship of historical category complexity and predicted accuracy.