To enhance the spatio-temporal modeling capability of traffic flow prediction, a prediction method based on an adaptive graph attention gated recurrent unit network (AGA-GRU) was proposed. The method constructed a learnable adjacency matrix and obtains a probabilistic structure through softplus activation and row normalization. A row-wise Top-K retention and renormalization strategy was introduced to achieve hard sparsification, while an L1 sparsity regularization was applied to the structural parameters to improve interpretability and robustness. In the spatial dimension, multi-head graph attention (GAT) was employed to suppress noisy edges and highlight key correlations. In the temporal dimension, gate recurrent unit (GRU) was used to capture sequential dependencies. The two components were cascaded to form an efficient spatio-temporal modeling framework. The training objective combined prediction error and sparsity regularization. Experiments conducted on real-world datasets, with autoregressive integrated moving average model (ARIMA), bidirectional long short-term memory network (Bi-LSTM), graph neural network-long short-term memory network (GNN-LSTM), spatio-temporal graph convolutional network (ST-GCN), and multi-task graph neural network (MTGNN) as baseline models, were evaluated using mean absolute error (MAE), mean squared error (MSE), and root mean squared error (RMSE) metrics. The results show that AGA-GRU outperforms the compared methods in multi-step forecasting tasks and learns sparse and interpretable graph structures (e.g., self-loops and a few key cross-link connections), achieving a good balance between prediction accuracy and model interpretability. This method is well-suited for traffic flow prediction in urban road networks.
MTGNN融合了图神经网络与时间卷积机制,在短期预测中取得了较优结果。例如,在济南数据集的 15 min 预测中,MTGNN的MAE为 0.475 5,仅次于 AGA-GRU。然而,由于其图结构仍依赖预定义邻接关系,模型的泛化能力在跨场景应用中受到一定限制。
相比之下,所提出的AGA-GRU在两个数据集和3个时间尺度上均展现出稳定且优越的性能。在杭州数据集上,AGA-GRU在 15、30、45 min的预测中均取得最低或接近最低的MAE与RMSE,其中 15 min MAE为2.446 6,优于所有对比模型。在济南数据集上,其优势表现得更为明显,15 min MAE达到0.472 1,RMSE为0.873 5,MAPE为18.18%,均为最佳结果,并在30 min和45 min 预测中保持领先。该结果表明,AGA-GRU 在捕捉复杂交通流的动态时空依赖关系方面更具优势,能够有效适应不同城市、不同时间尺度的交通预测任务。
为验证每个关键模块的作用,包含以下消融配置并在同一数据口径下比较(以30 min MAE、RMSE、MAPE为主):①AGA-GRU(full):完整模型(自适应邻接+GAT +GRU+L1正则);②w/o自适应图:邻接使用静态 AA 或单位矩阵(仅自环);③w/o注意力(用 GCN 替代 GAT);④w/o L1稀疏正则:λsparse=0;⑤Top-K变化:Top-K在{1,3,5,10,12,15}上的性能比较。
XiaoJianli, QiuXue, ZhangYang, et al. Review on transportation large models[J]. Journal of Traffic and Transportation Engineering, 2025, 25(1): 8-28.
[3]
ZhaoL, SongY, ZhangC, et al. T-GCN: a temporal graph convolutional network for traffic prediction[J]. IEEE transactions on intelligent transportation systems, 2019, 21(9): 3848-3858.
[4]
SmithB L, WilliamsB M, OswaldR K. Comparison of parametric and nonparametric models for traffic flow forecasting[J]. Transportation Research Part C: Emerging Technologies, 2002, 10(4): 303-321.
[5]
TianX, WuS, XingX, et al. Traffic flow prediction based on improved deep extreme learning machine[J]. Scientific Reports, 2025, 15(1): 7421.
[6]
ChenY T, LiuA, LiC, et al. Traffic flow prediction based on spatial-temporal multi factor fusion graph convolutional networks[J]. Scientific Reports, 2025, 15(1): 12612.
[7]
HussainB, AfzalM K, AhmadS, et al. Intelligent traffic flow prediction using optimized GRU model[J]. IEEEAccess, 2021, 9: 100736-100746.
[8]
MaX, TaoZ, WangY, et al. Long short-term memory neural network for traffic speed prediction using remote microwave sensor data[J]. Transportation Research Part C: Emerging Technologies, 2015, 54: 187-197.
YaoJunfeng, HeRui, ShiTongtong, et al. Review on traffic flow prediction methods based on machine learning[J]. Journal of Tra-ffic and Transportation Engineering, 2023, 23(3): 44-67.
[15]
BaoY, HuangJ, ShenQ, et al. Spatial-temporal complex graph convolution network for traffic flow prediction[J]. Engineering Applications of Artificial Intelligence, 2023, 121: 106044.
HuZuoan, DengJincheng, HanJinli, et al. Review on application of graph neural network in traffic prediction[J]. Journal of Traffic and Transportation Engineering, 2023, 23(5): 39-61.
[18]
ConcasA, FenuC, ReichelL, et al. Chained structure of directed graphs with applications to social and transportation networks[J]. Applied Network Science, 2022, 7: 64.
[19]
GhalmaneZ, DaoudB. Graph-based learning for multimodal route recommendation[C]// 10th International Conference on Internet of Things, Big Data and Security. Porto: Institute for Systems and Technologies of Information, Control and Communication, 2025: 313-320.
[20]
RahmaniS, BaghbaniA, BouguilaN, et al. Graph neural networks for intelligent transportation systems: a survey[J]. IEEE Transactions on Intelligent Transportation Systems, 2023, 24(8): 8846-8885.
[21]
YuB, YinH, ZhuZ. Spatio-temporal graph convolutional networks: a deep learning framework for traffic forecasting[J]. arXiv Preprint, arXiv: 1709.04875, 2017.
LiF, FengJ, YanH, et al. Dynamic graph convolutional recurrent network for traffic prediction: benchmark and solution[J]. ACM Transactions on Knowledge Discovery from Data, 2023, 17(1): 1-21.
[24]
WuZ, PanS, LongG, et al. Connecting the dots: multivariate time series forecasting with graph neural networks[C]// Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. New York: ACM, 2020: 753-763.
JiaXingli, QuYuanhai, ZhuHaoran, et al. Review on spatio-temporal graph neural networks in traffic prediction: from model deconstruction to development path[J]. Journal of Traffic and Transportation Engineering, 2026, 26 (1): 46-74.
JiangTing, YangLiu, LiuYalin, et al. Traffic flow prediction based on decomposed dynamic spatio-temporal decomposition framework[J]. Science Technology and Engineering, 2025, 25(7): 3007-3017.
[29]
XuY, CaiX, WangE, et al. Dynamic traffic correlations based spatio-temporal graph convolutional network for urban traffic prediction[J]. Information Sciences, 2023, 621: 580-595.
[30]
WuZ, PanS, LongG, et al. Graph WaveNet for deep spatial-temporal graph modeling[J]. arXiv Preprint, arXiv: 1906.00121, 2019.
[31]
BaiL, YaoL, LiC, et al. Adaptive graph convolutional recurrent network for traffic forecasting[J]. Advances in Neural Information Processing Systems, 2020, 33: 17804-17815.
[32]
SongC, LinY, GuoS, et al. Spatial-temporal synchronous graph convolutional networks: a new framework for spatial-temporal network data forecasting[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Philadelphia: AAAI, 2020: 914-921.
[33]
ZhangG, ChenJ. Non-negative matrix factorization via adaptive sparse graph regularization[J]. Multimedia Tools and Applications, 2021, 80(8): 12507-12524.
[34]
ChenM, WangY, ChenY, et al. Adaptive graph representation for clustering[J]. IEEE Access, 2022, 10: 122981-122994.
[35]
ZhengH, LinF, FengX, et al. A hybrid deep learning model with attention-based conv-LSTM networks for short-term traffic flow prediction[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 22(11): 6910-6920.
[36]
ChenJ, ZhengL, HuY, et al. Traffic flow matrix-based graph neural network with attention mechanism for traffic flow prediction[J]. Information Fusion, 2024, 104: 102146.
[37]
ZhengC, FanX, WangC, et al. GMAN: a graph multi-attention network for traffic prediction[C]// Proceedings of the AAAI conference on artificial intelligence. Philadelphia: AAAI, 2020: 1234-1241.
[38]
HuoY, ZhangH, TianY, et al. A spatiotemporal graph neural network with graph adaptive and attention mechanisms for traffic flow prediction[J]. Electronics, 2024, 13(1): 212.
[39]
Al HamadaniR A J, MoslehM, Al-SallamiA H A, et al. Improvement of network traffic prediction in beyond 5G network using sparse decomposition and BiLSTM neural network[J]. Qubahan Academic Journal, 2025, 5(2): 156-176.
[40]
WangX, WuY, ZhuangD, et al. Low-rank Hankel tensor completion for traffic speed estimation[J]. IEEE Transactions on Intelligent Transportation Systems, 2023, 24(5): 4862-4871.
[41]
WeiH, ZhengG, GayahV, et al. A survey on traffic signal control methods[J]. arXiv Preprint, arXiv: 1904.08117, 2019.
[42]
WeiH, XuN, ZhangH, et al. Colight: learning network-level cooperation for traffic signal control[C]// Proceedings of the 28th ACM International Conference on Information and Knowledge Management. New York: ACM, 2019: 1913-1922.
[43]
ZhengG, XiongY, ZangX, et al. Learning phase competition for traffic signal control[C]// Proceedings of the 28th ACM International Conference on Information and Knowledge Management. New York: ACM, 2019: 1963-1972.