融合动态图递归与状态空间的时空耦合交通流预测模型

张红; 齐方正; 雒生俊; 张玺君; 侯亮; 黄海蓉

doi:10.11999/JEIT251198

融合动态图递归与状态空间的时空耦合交通流预测模型

doi: 10.11999/JEIT251198 cstr: 32379.14.JEIT251198

1.
兰州理工大学计算机与人工智能学院兰州 730050
2.
中南大学计算机学院长沙 410017

基金项目: 甘肃省重点人才项目(2024RCXM57)，甘肃省科技重大专项计划(25ZYJA037)，国家自然科学基金(62566036)

详细信息

作者简介:
张红：女，博士，副教授，研究方向为智能交通、图神经网络建模及运用、大数据分析

齐方正：男，硕士，研究方向为智能交通

雒生俊：男，研究方向为大数据分析

张玺君：男，博士，副教授，研究方向为大数据分析、智能交通

侯亮：男，硕士，讲师，研究方向为智能交通

黄海蓉：女，硕士，研究方向为智能交通

通讯作者:
张红　zhanghong@lut.edu.cn

中图分类号: TN911.7
计量
- 文章访问数: 155
- HTML全文浏览量: 92
- PDF下载量: 31
- 被引次数: 0
出版历程
- 收稿日期: 2025-11-13
- 修回日期: 2026-04-17
- 录用日期: 2026-04-17
- 网络出版日期: 2026-04-30

A Spatiotemporal Coupling Traffic Flow Prediction Model with Dynamic Graph Recursion and State Space

1.
School of Computer and Artificial Intelligence, Lanzhou University of Technology, Lanzhou 730050, China
2.
School of Computer Science and Engineering, Central South University, Changsha 410017, China

摘要

摘要: 准确的交通流预测是智能交通系统中的关键任务，其核心挑战在于如何有效捕捉城市路网中动态演化的空间结构以及复杂的时空相关关系。针对现有方法在建模交通路网的动态关联时难以自适应捕捉路网空间依赖特征，对空间特征表征能力有限，且计算效率低等问题，该文提出一种融合动态图递归与状态空间的时空交通流预测模型(DGGRU-Mamba)。该模型构建了时空嵌入生成器，将节点的空间位置信息与周期性时间特征联合编码，以增强图结构对交通流时间特征的感知能力；设计了动态图递归建模(DGRM)，通过多层动态图门控递归单元(DGGRU)动态构建邻接关系，捕捉路网交通状态演变引发的空间依赖性；建立了基于结构化状态转移机制的时空Mamba(ST-Mamba)，实现交通流的全局时序建模，在提升建模能力的同时降低计算开销。相较主流自注意力模型STAEformer和DGGRU-Mamba，所提模型在PEMS04数据集上的MAE, RMSE和MAPE分别降低约4.2%, 3.8%和2.9%，同时推理时间缩短约4.82 s，在提升预测精度的同时提高了计算效率。
- 交通流预测 /
- 动态图门控递归 /
- 状态空间转移 /
- 动态时空特征 /
- 多步预测
Abstract: Objective Accurate traffic flow prediction is a key task in intelligent transportation systems. However, it remains challenging to capture dynamically evolving spatial structures and complex spatiotemporal dependencies in urban road networks. To address these issues, this paper proposes DGGRU-Mamba, a spatiotemporal traffic flow prediction framework that integrates dynamic graph recurrent modeling with a structured state space mechanism. The model jointly captures dynamic spatial dependencies and long-range temporal dependencies. Methods DGGRU-Mamba contains two core modules: Dynamic Graph Recurrent Modeling (DGRM) and Spatiotemporal Mamba (ST-Mamba). A spatiotemporal embedding generator is first designed to jointly encode periodic temporal information and node-specific spatial features, thereby supporting adaptive graph construction. DGRM dynamically updates time-varying adjacency structures through Dynamic Graph Gated Recurrent Units (DGGRUs), which enables adaptive modeling of evolving spatial dependencies. ST-Mamba uses structured state transitions to efficiently capture long-range temporal dependencies. In addition, a dual-branch prediction scheme with Forecast and Backcast branches is used to improve multi-step prediction accuracy and reduce cumulative errors. Results and Discussions DGGRU-Mamba is evaluated on four benchmark datasets, namely PEMS03, PEMS04, PEMS07, and PEMS08. Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) are used as evaluation metrics. Experimental results show that DGGRU-Mamba achieves strong performance on all datasets. On PEMS04, compared with the mainstream attention-based model STAEformer, DGGRU-Mamba reduces MAE, RMSE, and MAPE by approximately 4.2%, 3.8%, and 2.9%, respectively. Its inference time is also shortened by 4.82 s. These results indicate that the proposed framework improves prediction accuracy while maintaining high computational efficiency. The performance gains mainly arise from the complementary effects of DGRM and ST-Mamba, which strengthen dynamic spatial dependency modeling and long-range temporal dependency learning with lower computational cost. Conclusions This paper proposes DGGRU-Mamba, a spatiotemporal traffic flow prediction framework for modeling dynamic spatial structures and long-range temporal dependencies in complex traffic networks. By integrating dynamic graph recurrent modeling with a structured state space mechanism, the framework achieves a favorable balance between prediction accuracy and computational efficiency. Experiments on multiple benchmark datasets verify its effectiveness and scalability in multi-step traffic flow prediction. Future work will consider external factors, such as weather and traffic events, to further improve its applicability in real traffic scenarios.
- Traffic flow prediction /
- Dynamic graph gated recurrence /
- state space model /
- Spatiotemporal dependencies /
- Multi-step prediction

HTML全文

图 1 融合动态图递归与状态空间的时空交通流预测模型(DGGRU-Mamba)

下载: 全尺寸图片幻灯片

图 2 DGRM中的动态图结构构建示意图

下载: 全尺寸图片幻灯片

图 3 动态图门控递归单元结构图

下载: 全尺寸图片幻灯片

图 4 时空Mamba(ST-Mamba)结构示意图

下载: 全尺寸图片幻灯片

图 5 数据集PEMS04上预测误差随预测步长变化趋势

下载: 全尺寸图片幻灯片

图 6 DGGRU-Mamba在PEMS04, PEMS08数据集上的消融实验误差比较

下载: 全尺寸图片幻灯片

图 7 DGGRU-Mamba在3个数据集上的训练损失曲线、验证损失曲线与测试损失曲线

下载: 全尺寸图片幻灯片

图 8 不同嵌入维度设置下DGGRU-Mamba模型在PEMS04与PEMS08数据集上的MAE与RMSE表现

下载: 全尺寸图片幻灯片

表 1 数据集详细信息

数据集	节点	时间跨度	Samples	步长(min)	类型
PEMS03	358	2018.09–2018.11	26,208	5	流量
PEMS04	307	2018.01–2018.02	16,992	5	流量
PEMS07	883	2017.05–2017.08	28,224	5	流量
PEMS08	170	2016.07–2016.08	17,856	5	流量

下载: 导出CSV

表 2 对比实验

模型	PEMS03			PEMS04			PEMS07			PEMS08
模型	MAE	RMSE	MAPE(%)	MAE	RMSE	MAPE(%)	MAE	RMSE	MAPE(%)	MAE	RMSE	MAPE(%)
STGCN	17.55	30.42	17.34	21.16	34.89	13.83	25.33	39.34	11.21	17.50	27.09	11.29
AGCRN	15.98	28.25	15.23	19.27	32.26	12.92	22.27	36.55	9.12	19.05	25.22	9.54
DCRNN	17.99	30.31	18.34	21.22	33.44	14.17	25.22	38.61	11.82	16.82	26.36	10.92
GWN	19.12	32.77	18.89	24.89	39.66	17.29	26.39	41.50	11.97	18.28	30.05	12.15
STG-NCDE	15.59	26.68	14.99	19.42	31.34	12.98	20.78	35.00	8.95	16.34	25.51	10.26
STGOOE	15.56	26.64	14.96	18.91	30.32	12.91	20.32	35.43	8.90	16.34	25.44	10.58
ASTGCN(r)	17.34	29.56	17.21	22.92	35.22	16.56	24.01	37.87	10.73	18.25	28.06	11.64
STNorm	15.32	25.93	14.37	18.96	30.98	12.69	20.50	34.66	8.75	15.41	24.77	9.76
GMAN	16.87	27.92	18.23	19.14	31.60	13.19	20.97	34.10	9.05	15.31	24.92	10.13
STAEformer	15.05	25.55	14.91	18.85	30.85	12.55	20.25	34.20	8.65	15.10	24.50	9.92
PDFormer	14.85	25.35	14.75	18.65	30.70	12.45	20.05	33.95	8.55	14.91	24.20	9.75
DGGRU-Mamba	14.57	25.06	14.61	18.37	30.49	12.29	19.75	33.40	8.31	14.51	23.86	9.45

下载: 导出CSV

表 3 消融实验

模型	PEMS04			PEMS08
模型	MAE	RMSE	MAPE(%)	MAE	RMSE	MAPE(%)
w/o DGRM	18.85	31.39	12.74	15.14	24.92	10.01
w/o Backcast	18.75	31.19	12.64	15.04	24.72	9.91
w/o ST-Mamba	19.02	31.59	12.81	15.31	25.12	10.08
w/o STE	18.66	31.04	12.55	14.95	24.57	9.82
DGGRU-Mamba	18.37	30.49	12.29	14.51	23.86	9.45

下载: 导出CSV

表 4 训练与推理时间

数据、时间、模型		STG-NCDE	STSGCN	AGCRN	STGODE	GWN	STAEformer	PDFormer	DGGRU-Mamba
PEMS04	训练时间(s/轮)	118.6	181.9	6.5	35.2	32.18	92.61	108.32	21.37
PEMS04	推理时间(s)	12.28	4.63	1.12	4.08	3.37	6.42	7.08	1.58
PEMS08	训练时间(s/轮)	43.2	61.28	3.9	22.3	10.64	74.22	86.85	17.99
PEMS08	推理时间(s)	4.3	12.4	0.5	2.1	1.2	5.1	5.8	1.7

下载: 导出CSV

参考文献(16)

[1]	张红, 伊敏, 张玺君, 等. 长期Transformer和自适应傅里叶变换的动态图卷积交通流预测研究[J]. 电子与信息学报, 2025, 47(7): 2249–2262. doi: 10.11999/JEIT241076. ZHANG Hong, YI Min, ZHANG Xijun, et al. Long-term Transformer and adaptive Fourier transform for dynamic graph convolutional traffic flow prediction study[J]. Journal of Electronics & Information Technology, 2025, 47(7): 2249–2262. doi: 10.11999/JEIT241076.
[2]	智慧, 段苗苗, 杨利霞, 等. 一种基于区块链和联邦学习融合的交通流预测方法[J]. 电子与信息学报, 2024, 46(9): 3777–3787. doi: 10.11999/JEIT240030. ZHI Hui, DUAN Miaomiao, YANG Lixia, et al. A traffic flow prediction method based on the fusion of blockchain and federated learning[J]. Journal of Electronics & Information Technology, 2024, 46(9): 3777–3787. doi: 10.11999/JEIT240030.
[3]	殷礼胜, 刘攀, 孙双晨, 等. 基于互补集合经验模态分解和改进麻雀搜索算法优化双向门控循环单元的交通流组合预测模型[J]. 电子与信息学报, 2023, 45(12): 4499–4508. doi: 10.11999/JEIT221172. YIN Lisheng, LIU Pan, SUN Shuangchen, et al. Traffic flow combined prediction model based on complementary ensemble empirical mode decomposition and bidirectional gated recurrent unit optimized by improved sparrow search algorithm[J]. Journal of Electronics & Information Technology, 2023, 45(12): 4499–4508. doi: 10.11999/JEIT221172.
[4]	ZHANG Qingyong, LI Changwu, SU Fuwen, et al. Spatiotemporal residual graph attention network for traffic flow forecasting[J]. IEEE Internet of Things Journal, 2023, 10(13): 11518–11532. doi: 10.1109/JIOT.2023.3243122.
[5]	LI Biyue, LI Zhishuai, CHEN Jun, et al. MAST-GNN: A multimodal adaptive spatio-temporal graph neural network for airspace complexity prediction[J]. Transportation Research Part C: Emerging Technologies, 2024, 160: 104521. doi: 10.1016/j.trc.2024.104521.
[6]	LI Zhuolin, ZHANG Gaowei, YU Jie, et al. Dynamic graph structure learning for multivariate time series forecasting[J]. Pattern Recognition, 2023, 138: 109423. doi: 10.1016/j.patcog.2023.109423.
[7]	WU Di, PENG Kai, WANG Shangguang, et al. Spatial–temporal graph attention gated recurrent transformer network for traffic flow forecasting[J]. IEEE Internet of Things Journal, 2024, 11(8): 14267–14281. doi: 10.1109/JIOT.2023.3340182.
[8]	LIU Shipeng and WANG Xingjian. An improved transformer based traffic flow prediction model[J]. Scientific Reports, 2025, 15(1): 8284. doi: 10.1038/s41598-025-92425-7.
[9]	SHAO Zhiqi, WANG Ze, YAO Xusheng, et al. ST-MambaSync: Complement the power of mamba and transformer fusion for less computational cost in spatial-temporal traffic forecasting[J]. Information Fusion, 2025, 117: 102872. doi: 10.1016/j.inffus.2024.102872.
[10]	LI Yaguang, YU R, SHAHABI C, et al. Diffusion convolutional recurrent neural network: Data-driven traffic forecasting[C]. The 6th International Conference on Learning Representations, Vancouver, Canada, 2018: 1–16.
[11]	GUO Shengnan, LIN Youfang, FENG Ning, et al. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting[C]. The 33rd AAAI Conference on Artificial Intelligence, Honolulu, USA, 2019: 922–929. doi: 10.1609/aaai.v33i01.3301922.
[12]	YE Yaqin, XIAO Yue, ZHOU Yuxuan, et al. Dynamic multi-graph neural network for traffic flow prediction incorporating traffic accidents[J]. Expert Systems with Applications, 2023, 234: 121101. doi: 10.1016/j.eswa.2023.121101.
[13]	ZHENG Chuanpan, FAN Xiaoliang, WANG Cheng, et al. GMAN: A graph multi-attention network for traffic prediction[C]. The 34th AAAI Conference on Artificial Intelligence, New York, USA, 2020: 1234–1241. doi: 10.1609/aaai.v34i01.5477.
[14]	DENG Jinliang, CHEN Xiusi, JIANG Renhe, et al. ST-norm: Spatial and temporal normalization for multi-variate time series forecasting[C]. The 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, Singapore, 2021: 269–278. doi: 10.1145/3447548.3467330.
[15]	SHAO Zhiqi, BELL M G H, WANG Ze, et al. St-mamba: Spatial-temporal selective state space model for traffic flow prediction[EB/OL]. https://arxiv.org/abs/2404.13257, 2024.
[16]	CHOI J, KIM H, AN M, et al. SpoT-mamba: Learning long-range dependency on spatio-temporal graphs with selective state spaces[C]. The 3rd International Workshop on Spatio-Temporal Reasoning and Learning (STRL 2024) co-located with the 33rd International Joint Conference on Artificial Intelligence, Jeju island, Korea, 2024.