Short-term traffic flow prediction is a key technology for intelligent transportation systems to achieve efficient management and real-time regulation. To address the shortcomings of existing methods in spatio-temporal dependency representation and error accumulation, a dual-path spatio-temporal graph attention network model was proposed to achieve more accurate and robust short-term traffic flow prediction. Firstly, a dual-path graph attention mechanism for both static and dynamic conditions was designed to extract spatial features. In the static attention branch, stable spatial relationships were built based on the physical topology of the road network, while in the dynamic attention branch, sudden spatial correlations were adaptively captured according to real-time traffic flow data. The two branches were weighted and fused through a learnable fusion parameter, and the influence between static topology and dynamic state was adaptively balanced, so that comprehensive spatial features were extracted. Secondly, a long short-term memory network encoder structure was adopted to hierarchically extract and flexibly map the features of time series. Finally, the decoder introduced a dynamic attenuation hybrid teaching-learning mechanism, integrating recursive and teaching decoding modes to predict traffic flow, in order to effectively prevent error accumulation. The PEMS04 and PEMS08 datasets were used for training, validation and testing. The results show that the model achieves approximately 4.23%, 3.18%, 6.90% and 3.89%, 2.59%, 6.44% improvements over the optimal spatio-temporal comparison model convolutional neural network-long short-term memory network(CNN-LSTM) in terms of mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE) indicators respectively. It can be seen that the model is capable of effectively balancing the static topology and dynamic state of the road network, and significantly reduces error accumulation through the hybrid teaching-learning mechanism, demonstrating its superior performance and application potential in short-term traffic flow prediction.
实验在PEMS08、PEMS04两个公开交通数据集上测试,这两个数据集均来自美国加利福尼亚州交通局(california department of transportation,Caltrans)的交通性能监测系统(caltrans performance measurement system,PEMS)(http://pems.dot.ca.gov),包含该州2个地区的真实高速公路数据。具体来看,道路检测器每30 s采集一次原始数据,再以5 min为时间间隔进行聚合处理,因此一天的交通数据流中共包含288个数据点;所采集的交通特征涵盖交通流量、道路占用率与速度,所有实验均采用这3个核心特征,且为了保留交通流数据中的真实波动和突变信息,从而体现模型结构本身在处理原始、复杂、高波动性数据上的鲁棒性和有效性,没有对数据集进行额外预处理,数据集的具体信息如表1所示。
3.2 评价指标
选定3个广泛使用的流量预测性能指标进行不同模型下预测准确性的对比评估,包括平均绝对误差(mean absolute error,MAE)、均方根误差(root mean square error,RMSE)、平均百分比误差(mean absolute percentage error,MAPE),其中MAE反映模型整体偏差稳定性,RMSE揭示模型对极端值的鲁棒性,MAPE说明不同流量水平的适应性。这些评估指标越小表示模型预测性能越好。计算公式分别为
CaoQianxia, MaiQianlong, LüSongtao, et al. Road network traffic forecasting based on short-term nonlinear perturbation adaptive graph[J]. China Journal of Highway and Transport, 2025, 38(7): 275-288
ZhangJin, PiYu, SunCheng, et al. Spatial-temporal encoder-decoder model for traffic flow prediction[J]. Journal of National University of Defense Technology, 2025, 47(3): 173-182.
[5]
SattarzadehA R, KutadinataR J, PathiranaP N, et al. A novel hybrid deep learning model with ARIMA Conv-LSTM networks and shuffle attention layer for short-term traffic flow prediction[J]. Transportmetrica A: Transport Science, 2025, 21(1): 2236724.
[6]
DongH, ZhangH, DingF, et al. Co-evolving traffic state parameters prediction based on mechanism-data blending driven deep learning[J]. IEEE Transactions on Intelligent Transportation Systems, 2025, 26: 3084-3100.
[7]
DuanG, DuY, ShangY, et al. Research on support vector regression short-time traffic flow prediction model for secondary roads based on associated road analysis[J]. Applied Sciences, 2025, 15(4): 1779.
WangFei, SunPengfei. Short term forecasting of air traffic flow based on decomposition and integration method[J]. Science Technology and Engineering, 2021, 21(35): 15270-15276.
[10]
LinG, LinA, GuD. Using support vector regression and k-nearest neighbors for short-term traffic flow prediction based on maximal information coefficient[J]. Information Sciences, 2022, 608: 517-531.
TangZehui, ZhaoDan, WangShengyou. Short-term traffic flow prediction based on CNN-LSTM-AM[J]. Science Technology and Engineering, 2024, 24(31): 13562-13567.
[13]
MondalM A, RehenaZ. Stacked lstm for short-term traffic flow prediction using multivariate time series dataset[J]. Arabian Journal for Science and Engineering, 2022, 47(8): 10515-10529.
[14]
LiuY, FengT, RasouliS, et al. ST-DAGCN: a spatiotemporal dual adaptive graph convolutional network model for traffic prediction[J]. Neurocomputing, 2024, 601: 128175.
[15]
LiF, FengJ, YanH, et al. Dynamic graph convolutional recu-rrent network for traffic prediction: benchmark and solution[J]. ACM Transactions on Knowledge Discovery from Data, 2023, 17(1): 1-21.
[16]
LiuJ, WangY, ZhuJ, et al. Amultilayer spatiotemporal correlation-aware graph attention network for traffic flow prediction[J]. IEEE Transactions on Neural Networks and Learning Systems, 2025, 36: 30903.
MengXiangfu, XuYongjie, WengXue, Spatio-temporal cross-attention feature fusion model for traffic flow prediction[J]. Journal of Frontiers of Computer Science and Technology, 2025, 19(7): 1931-1944.
ZhongSonghao, LuYi, LiLi, et al. Long term traffic flow forecasting model based on cross period sparse technique[J]. Science Technology and Engineering, 2025, 25(24): 10468-10477.
[21]
HeG, HuangW, ZhuY, et al. Adaptive spatial-temporal dependence graph convolution neural network for traffic flow prediction[J]. Expert Systems with Applications, 2025, 281: 127564.
[22]
ChoiY W, ShinH B, LeeS W. Brain-guided self-paced curriculum learning for adaptive human-machine interfaces[J]. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2025, 55: 4693-4704.
[23]
ChenH, HuangJ, LuY, et al. Urban traffic flow prediction based on multi-spatio-temporal feature fusion[J]. Neurocomputing, 2025, 638: 130117.
[24]
JiangJ, HanC, ZhaoW X, et al. PDFormer: propagation delay-aware dynamic long-range transformer for traffic flow prediction[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Philadelphia: AAAI, 2023: 4365-4373.
[25]
FanJ, WengW, TianH, et al. RGDAN: a random graph diffusion attention network for traffic prediction[J]. Neural Networks, 2024, 172: 106093.
[26]
LanS, MaY, HuangW, et al. DSTAGNN: dynamic spatial-temporal aware graph neural network for traffic flow forecasting[C]// International Conference on Machine Learning. Cambridge, MA: PMLR, 2022: 11906-11917.
[27]
LuoM, DouH, ZhengN. Spatiotemporal prediction of urban traffics based on deep GNN[J]. Computers, Materials and Continua, 2024, 78(1): 265-282.
[28]
JinM, ZhengY, LiY F, et al. Multivariate time series forecasting with dynamic graph neural odes[J]. IEEE Transactions on Knowledge and Data Engineering, 2022, 35(9): 9168-9180.
[29]
ChoiJ, ChoiH, HwangJ, et al. Graph neural controlled differential equations for traffic forecasting[C]// Proceedings of the AAAI Conference on Artificial Intelligence. Philadelphia: AAAI, 2022: 6367-6374.
[30]
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.
[31]
SherafatE, FarooqB, KarbasiA H, et al. Attention-LSTM for multivariate traffic state prediction on rural roads[J]. arXiv Preprint, arXiv: 2301.02731, 2023.
HeWenwu, PeiBoyu, LiYating. Traffic flow forecasting based on bi-directional adaptive gating graph convolutional networks[J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(1): 187-197.