UPSI Digital Repository (UDRep)
|
UPSI Digital Repository (UDRep)
|
|
|
|
||||||||||||||||||||||||
| Abstract : Perpustakaan Tuanku Bainun |
| With the acceleration of urbanization, metro passenger flow patterns in megacities have become increasingly complex. Shanghai, with the world_s largest metro network, poses significant challenges to accurate demand forecasting. Traditional approaches often neglect passenger behavioral heterogeneity and service satisfaction, resulting in limited prediction performance. This study integrates travel behavior mining and satisfaction perception into forecasting by proposing a Context-Aware Graph Transformer (CaGT) model that fuses spatio-temporal features with human-centered factors.The dataset comprises smart card transaction records (5-minute intervals), network topology of more than 500 stations, and 2660 valid passenger satisfaction surveys based on the SERVQUAL framework. Analysis revealed that peak-hour demand is highly concentrated, with transfer bottlenecks at People_s Square and Century Avenue, and five major commuting corridors were identified. Survey results showed that crowding, convenience transfer, and in-train comfort significantly affect satisfaction and loyalty.The CaGT model integrates Graph Attention Networks (GAT) with Transformer encoders, embedding contextual factors such as satisfaction, crowding index, and temporal preferences into dynamic graph weights. Experimental evaluations demonstrated that CaGT significantly outperformed benchmark models (MLP, LSTM, STGCN, ASTGCN), achieving RMSE = 10.21, MAE = 6.48, and MAPE = 8.7%, with superior accuracy particularly during peak hours.This study contributes theoretically by bridging passenger behavior with graph-based forecasting, and practically by offering actionable insights for metro scheduling, transfer management, and passenger service improvement. The framework holds strong potential for cross-city transferability and multimodal transit integration in the future. |
| This material may be protected under Copyright Act which governs the making of photocopies or reproductions of copyrighted materials. You may use the digitized material for private study, scholarship, or research. |