The aim is to develop a framework that accurately predicts short-term passenger flow in urban subway systems by addressing dynamic station characteristics and multiscale dependencies.
Developed a multi-graph convolutional-convolutional network (SF-MGCN-CNN) integrating station features.
Validated the framework using data from the Chengdu subway network with performance compared to baseline methods.
Conducted ablation studies to analyze contributions of individual framework components.
The SF-MGCN-CNN demonstrated superior accuracy in passenger flow prediction over baseline methods.
Ablation studies confirmed the importance of integrating station characteristics and spatiotemporal features for improved predictions.