Summary
In this chapter, we began by recognizing that robust trajectory analysis is essential for applications ranging from deliveries and epidemic forecasting to city-scale planning. Guided by the innovative approach outlined in Tang, P., Yang, C., Xing, T., Xu, X., Jiang, R., and Sezaki, K. (2024), we emphasized the transformative potential of text-based LLMs for mobility prediction. Their framework directed our design of a method capable of intelligently filling gaps in real-time synthetic datasets through carefully structured prompts.
We then built a Python-based trajectory simulator that randomizes movement on a grid, mirroring typical user paths. It assigns day and timeslot indices, which enabled us to capture the temporal aspect of mobility. Critically, we inserted synthetic gaps marked as 999, 999, approximating real-world data dropouts or missing logs. Next, we integrated an orchestrator function that adds instructions with this synthetic data before directing them to...
