TrajectoryGeneration-Act2Loc

Human mobility data play a crucial role in many fields such as infectious diseases, transportation, and public safety. Although the development of information and communication technologies has made it easy to collect individual-level positioning records, raw individual trajectory data are still limited in availability and usability due to privacy issues. Developing models to generate synthetic trajectories that are as realistic as possible while preserving privacy is a promising solution. This study proposes a novel trajectory generation method called <strong>Act2Loc</strong> (<strong>Act</strong>ivity <strong>to</strong> <strong>Loc</strong>ation), which combines machine learning and mechanistic models. First, an activity-sequence generation model is constructed based on machine learning (i.e., K-medoids and LSTM) to generate activity sequences of a group of individuals that are representative of the temporal patterns of human daily activities. Then, a spatial-location selection model is proposed based on mechanistic models (e.g., unified opportunity model) to explicitly determine the specific locations of the activities in each generated sequence. Experimental results show that compared to the baselines, trajectories generated by Act2Loc can better reproduce the spatio-temporal characteristics of the real data, proving its potential for generating high-quality synthetic trajectories in practice. This research offers new insights on the development of knowledge-guided GeoAI models for human mobility.