Short-Term Metro Passenger Flow Forecasting Using a Context-Aware Graph Transformer

DOI: https://doi.org/10.62517/jbdc.202601329

Author(s)

Ruilin Li1, Honghong Li2,*, Jianhua Wu2, Yongfu He1

Affiliation(s)

1Gongqing College of Nanchang University, Jiujiang, Jiangxi, China 2School of Electronic and Information Engineering, Nanchang University, Nanchang, Jiangxi, China *Corresponding Author

Abstract

Metro operators require dependable station-level forecasts over the next hour to align service capacity with demand and intervene before local crowding develops. Such forecasts are difficult because recent flow trajectories interact with inter-station movement and with disruptions caused by weather, holidays, and events. This paper develops a Context-Aware Graph Transformer with Adaptive Context Fusion (CaGT). The spatial block learns unequal influence among connected stations, while the temporal block represents the evolution of the network over successive 15-min intervals. Instead of appending every external variable directly to the historical series, a learned gate controls how much contextual information enters the prediction at each state. Tests on operational records from all 94 stations of Nanchang Metro produced an RMSE of 1.76, an MAE of 1.28, and a MAPE of 6.9. Relative to the Transformer baseline, RMSE and MAE decreased by 20.4% and 21.0%, respectively. The ablation results attribute these gains to both attention-weighted station interaction and state-dependent context use. CaGT therefore offers a practical means of representing short-horizon metro demand when routine temporal regularity is interrupted by changing operating conditions.

Keywords

Short-Term Ridership Prediction; Metro Networks; Graph Transformer; Traph Attention; Adaptive Context Fusion; Spatio-Temporal Modelling

References

[1] Li, Y., Yu, R., Shahabi, C., &Liu, Y. (2018). Diffuion convolutional recurrent neural network:Data-driven traffic forecasting. International Conference on Learning Representations (ICLR). [2] Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit,J.,Jones,L.,Gomez,A.N.,Kaiser,Ł,&Polosukhin,I.(2017). Attention Is All You Need.Advances in Neural Information Processing Systems, 30, 5998–6008. [3] Hu, S., Chen, J., Zhang, W., Liu, G., &Chang, X. (2024). Graph transformer embedded deep learning for short-term passenger flow prediction in urban rail transit systems:A multi-gate mixture-of -experts model.Information Sciences, 679,121095. [4] Jiang, J., Han, C., Zhao, W.X., &Wang,J. (2023).PDFormer:PropagationDelay-Aware Dynamic Long-Range Transformer for Traffic Flow Prediction. Proceedings of the AAAI Conference on Artificial Intelligence, 37(4), 4365– 4373. [5] Yu, B., Yin, H., &Zhu, Z. (2018). Spatio-temporal graph convolutional networks: A deep learning framework for traffic forecasting.Proceedings of the 27th International Joint Conference on Artificial Intelligence (IJCAI), 3634– 3640. [6] Veličković, P., Cucurull, G., Casanova, A., Romero, A., Liò, P., &Bengio, Y. (2018).Graph Attention Networks. International Conference on Learning Representations (ICLR). [7] Guo, S., Lin, Y., Feng, N., Song, C., &Wan, H. (2019). Attention based spatial- temporal graph convolutional networks for traffic flow forecasting. Proceedings of the AAAI Conference on Artificial Intelligence,33(1),922–929. [8] Wu, Z., Pan, S., Long, G., Jiang, J., &Zhang, C. (2019). Graph WaveNet for deep spatial-temporal graph modelling. Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI), 1907–1913. [9] Zheng, C., Fan, X., Wang, C., &Qi, J. (2020) GMAN: A Graph Multi-Attention Network for Traffic Prediction. Proceedings of the AAAI Conference on Artificial Intelligence, 34(1), 1234– 1241. [10] Jin, D., Shi, J., Wang, R., Li, Y., Huang, Y., &Yang, Y. (2023). Trafformer: Unify Time and Space in Traffic Prediction. Proceedings of the AAAI Conference on Artificial Intelligence, 37(7), 8114 – 8122. [11] Hu, Z., Chang, S., Zhang, W., et al. (2026). STKGformer: A spatio-temporal knowledge graph enhanced transformer for passenger flow prediction. Information Sciences, 735, 123055. [12] Zhan, S., Cai, Y., Xiu, C., Zuo, D., Wang, D., &Wong, S.C.(2024).Parallel framework of a multi-graph convolutional network and gated recurrent unit for spatial- temporal metro passenger flow prediction. Expert Systems with Applications, 251,123982. [13] Chen,P.,Fu,X.,&Wang,X.(2022).A graph convolutional stacked bidirec- tional LSTM neural network for metro ridership prediction.IEEE Transactions on Intelligent Transportation Systems, 23(7), 6950–6962. [14] Bao, J., Kang, J., Yang, Z., & Chen, X. (2022). Forecasting network-wide multi-step metro ridership with an attention- weighted multi-view graph to sequence learning approach. Expert Systems with Applications,210,118475. [15] Fu, X., Zuo, Y., Wu, J., Yuan, Y., &Wang, S. (2022). Short-term prediction of metro passenger flow with multi-source data: A neural network model fusing spatial and temporal features. Tunnelling and Underground Space Technology, 124,104486. [16] He, Y., Li, L., Zhu, X., &Tsui, K.L. (2022). Multi-Graph Convolutional-Recurrent Neural Network for short-term forecasting of transit passenger flow. IEEE Transactions on Intelligent Transportation Systems, 23(10), 18155–18174. [17] Zeng, J., &Tang, J. (2023).Combining knowledge graph into metro passenger flow prediction: A split-attention relational graph convolutional network. Expert Systems with Applications, 213,118790. [18] Huang, H., Mao, J., Kang, L., Lu, W., Zhang, S., &Liu, L. (2024).A dynamic graph deep learning model with multivariate empirical mode decomposition for network-wide metro passenger flow prediction. Computer-Aided Civil and Infrastructure Engineering,39, 2596– 2618. [19] Li, P., Wang, S., Zhao, H., Yu, J., Hu, L., Yin, H., &Liu, Z.(2023). IG-Net: An interaction graph network model for metro passenger flow forecasting. IEEE Transactions on Intelligent Transportation Systems,24(4), 4147– 4157. [20] Wu, F., Zheng, C., Zhou, S., Lu, Y., Wu, Z., &Zheng, S.(2025).An interpretable approach to passenger flow prediction and irregular passenger travel pattern understanding in metro systems. Expert Systems with Applications,265, 125991. [21] Ren, Y., Tang, J., Ying, C., & Dong, Q. (2026). A multi-source feature- integrated spatiotemporal graph transformer network for short-term origin–destination passenger flow prediction in urban rail transit. Journal of Intelligent Transportation Systems, 1–19.2669875 [22] Yang, Q., Yu, J., Wang, Z., Hu, X., Chen, Y., & Li, F. (2026). Short-term metro flow forecasting via spatiotem- poral joint self-attention and multi- graph weighted fusion. International Journal of Transportation Science and Technology. Volume 159,106491,