PHGNet: Prototype-Guided Hypergraph for Traffic Forecasting

Researchers propose PHGNet, a spatiotemporal forecasting framework that uses prototype-guided hypergraph construction to capture high-order interactions among nodes with similar traffic patterns. The method addresses spatial heterogeneity in traffic systems by adaptively assigning pattern-similar nodes to hyperedges, enabling dynamic time-varying structures. A global-local node representation further improves reliability. The work is published on arXiv (2605.25554).