Mobility characterization based on spatio-temporal data

Spatio-temporal fingerprinting plays a crucial role in contemporary mobility analysis, particularly in understanding travel behaviors and inferring user identities from anonymized data. In this context, studies under the Trajectory-User Linking (TUL) framework have focused on developing various neural network models using Location-Based Social Network (LBSN) data. However, there is a notable gap in the literature regarding neural network models designed specifically to process sequential spatio-temporal data, such as GPS traces. To address this gap, this study introduces an innovative model called the Spatio-Temporal Sequential Graph Neural Network (STSeqGNN), capable of processing sequential spatio-temporal data effectively. The proposed model excels in leveraging both the graph structure of the map and the temporal dimension of the data, ensuring efficient and accurate processing. Additionally, it evaluates the mobility patterns and fundamental statistical properties of trajectories, enabling more precise predictions while preserving data integrity. The model's performance was evaluated on multiple datasets, and the results were assessed using the k-accuracy metric. The findings demonstrate that STSeqGNN achieves an accuracy of over 99\% across various datasets, highlighting its effectiveness in analyzing travel behaviors. These results underline the model's strong potential as a powerful tool for both theoretical exploration and practical applications in mobility analysis.