Forecasting & Urban Intelligence

I study forecasting and analysis methods for urban systems, focusing on lightweight yet effective architectures that support practical deployment.

Key Works

• PLGF (AAAI 2026): Improving fine-grained urban flow inference with lightweight architecture and focalized optimization.
• ExoST (arXiv 2025): A plug-and-play framework for exogenous-aware spatio-temporal forecasting.
• TimeEmb (NeurIPS 2025): Static-dynamic disentanglement for robust and efficient time series forecasting.

Research Focus

• Improve prediction accuracy under sparse or noisy urban observations.
• Model exogenous factors for stronger robustness and interpretability.
• Support real-time urban analytics and decision support with efficient models.