MM-STFlowNet: A Transportation Hub-Oriented Multi-Mode Passenger Flow Prediction Method via Spatial-Temporal Dynamic Graph Modeling

TL;DR

This work addresses the need for refined, multi-mode passenger flow forecasting in large transportation hubs by modeling inter-mode interactions through a dynamic spatial-temporal graph. The proposed MM-STFlowNet combines a temporal feature enhancement stage with a Spatial-Temporal Dynamic Graph Convolutional Recurrent Network (STDGCRN) and an adaptive channel attention module, followed by self-attention leveraging external factors to improve peak-period accuracy. Key contributions include a CEEMDAN-based series decomposition, a multi-channel temporal enhancement pipeline, a data-driven STDGCRN for joint spatial-temporal encoding, and the Enhanced Peak Exponential Loss (EPEL) to emphasize peak errors. Experiments on Guangzhounan Railway Station data (and a separate Traffic dataset) demonstrate state-of-the-art performance during peaks and holidays, highlighting the method's practical value for hub management and resource allocation.

Abstract

Accurate and refined passenger flow prediction is essential for optimizing the collaborative management of multiple collection and distribution modes in large-scale transportation hubs. Traditional methods often focus only on the overall passenger volume, neglecting the interdependence between different modes within the hub. To address this limitation, we propose MM-STFlowNet, a comprehensive multi-mode prediction framework grounded in dynamic spatial-temporal graph modeling. Initially, an integrated temporal feature processing strategy is implemented using signal decomposition and convolution techniques to address data spikes and high volatility. Subsequently, we introduce the Spatial-Temporal Dynamic Graph Convolutional Recurrent Network (STDGCRN) to capture detailed spatial-temporal dependencies across multiple traffic modes, enhanced by an adaptive channel attention mechanism. Finally, the self-attention mechanism is applied to incorporate various external factors, further enhancing prediction accuracy. Experiments on a real-world dataset from Guangzhounan Railway Station in China demonstrate that MM-STFlowNet achieves state-of-the-art performance, particularly during peak periods, providing valuable insight for transportation hub management.

Figures (15)

• Figure 1: Multi-mode refined passenger flow prediction for congestion avoidance. Source: adapted frompic1pic2pic3pic4.
• Figure 2: Analysis of spatial-temporal feature extraction and modeling based on the dynamic multi-mode traffic network. Source: adapted frompic1.
• Figure 3: The overall proposed framework for refined forecasting of multi-mode passenger flow.
• Figure 4: TCN. (a) displays a dilated causal convolution with dilation factors $d=1, 2, 4$ and filter size $k_T=2$.
• Figure 5: STDGCRU. Each memory unit has a data-driven dynamic graph convolution kernel. STDGCRN includes multi-layer STDGCRUs.