Heterogeneous graph traffic prediction considering spatial information around roads

Jiahui Chen, Lina Yang, Cang Qin, Yi Yang, Ling Peng, Xingtong Ge

Abstract: Highlights•A traffic prediction network HASTN based on heterogeneous graph is proposed.•Two meta-paths SRS and SGS are defined to build heterogeneous graphs.•The influence of road and geographical POI meta-paths on the model prediction performance is analyzed.•The effectiveness of HASTN is verified through experiments on traffic speed and traffic flow datasets.