Bi-directional diffusion dynamic graph convolutional network with multi-source information fusion and graph-level attention for traffic flow prediction

doi:10.11772/j.issn.1001-9081.2025070876

Abstract

Abstract:

Traffic data often exhibits significant dynamic spatial correlations and sudden pattern changes. Traditional models struggle to fully capture the spatio-temporal dependencies of real-world transportation networks. To address this challenge， a Bi-directional Diffusion Dynamic Graph convolutional Network with multi-source information fusion and graph-level attention for traffic flow prediction （BDDGNet） was proposed. The network integrated spatio-temporal information and multi-source dynamic features， utilizing a multi-head attention mechanism to construct dynamic graph adjacency relationships between nodes， thereby capturing spatial connectivity strength evolving over time in the traffic network. For spatial modeling， a bi-directional diffusion dynamic graph convolution module was designed to learn the propagation dynamics of node states in both directions， thereby enabling collaborative modeling of spatial dependencies in traffic flow. At the same time， a graph-level attention mechanism was introduced to extract global semantic information of key nodes， thereby enhancing the perception capability for the overall topological structure of the traffic network. Experimental results on two real-world traffic flow datasets， PEMSD4 and PEMSD8， showed that for 15-minute， 30-minute， and 60-minute prediction tasks， on the PEMSD4 dataset， BDDGNet achieved the Mean Absolute Error （MAE）， Root Mean Square Error （RMSE）， and Mean Absolute Percentage Error （MAPE） of 18.26， 30.26， and 12.16%， representing reductions of 25.2%， 19.3%， and 26.9% compared with Diffusion Convolutional Recurrent Neural Network （DCRNN） based on static graph structure， respectively； on the PEMSD8 dataset， BDDGNet achieved the MAE， RMSE， and MAPE of 14.07， 23.44， and 9.24%， representing reductions of 16.3%， 11.1%， and 15.4% compared with DCRNN， respectively. The proposed network can capture spatio-temporal dependencies in dynamic traffic environments more accurately and significantly enhance prediction accuracy.

Key words: traffic flow prediction, dynamic graph modeling, multi-source information fusion, multi-head attention mechanism, bi-directional diffusion dynamic graph convolution

摘要：

交通数据存在显著的动态空间相关性与突发性模式变化，传统模型难以全面建模真实交通网络的时空依赖特性。为了解决该问题，提出一种融合多源信息与图级注意力的双向扩散动态图卷积交通流预测网络（BDDGNet）。该网络融合了时空信息与多源动态特征，通过多头注意力机制构建节点间的动态图邻接关系，从而捕捉交通网络中随时间演化的空间连接强度。在空间建模方面，设计双向扩散动态图卷积模块，以从不同路径方向学习节点状态的传播动态，从而实现对交通流中空间依赖的协同建模；同时，引入图级注意力机制提取关键节点的全局语义信息，增强对整体交通网络拓扑结构的感知能力。在2个真实交通流数据集PEMSD4和PEMSD8上的实验结果表明， BDDGNet在15、30 和60 min预测任务中均取得了更优结果。综合平均结果显示：在PEMSD4数据集上， BDDGNet的平均绝对误差（MAE）、均方根误差（RMSE）和平均绝对百分比误差（MAPE）分别达到18.26、30.26和12.16%，相较于基于静态图结构的扩散卷积循环神经网络（DCRNN）分别降低了25.2%、19.3%和26.9%；在PEMSD8数据集上， BDDGNet的MAE、RMSE和MAPE分别达到14.07、23.44和9.24%，相较于DCRNN分别降低了16.3%、11.1%和15.4%。本文网络能够在多变的交通环境下更准确地刻画时空依赖关系，并显著提升预测精度。

关键词: 交通流预测, 动态图建模, 多源信息融合, 多头注意力机制, 双向扩散动态图卷积

CLC Number:

TP391.41

Jianqiang YAN, Beibei DONG, Boting QU, Chen PENG. Bi-directional diffusion dynamic graph convolutional network with multi-source information fusion and graph-level attention for traffic flow prediction[J]. Journal of Computer Applications, 2026, 46(7): 2327-2333.

颜建强, 董贝贝, 曲博婷, 彭晨. 融合多源信息与图级注意力的双向扩散动态图卷积交通流预测网络[J]. 《计算机应用》唯一官方网站, 2026, 46(7): 2327-2333.

Figures/Tables 5

References 28

[1]	Rani P， Sharma R. Intelligent transportation system for internet of vehicles based vehicular networks for smart cities ［J］. Computers and Electrical Engineering， 2023， 105： No.108543.
[2]	Lin C， Han G， Du J， et al. Spatiotemporal congestion-aware path planning toward intelligent transportation systems in software-defined smart city IoT ［J］. IEEE Internet of Things Journal， 2020， 7（9）： 8012-8024.
[3]	Ermagun A， Levinson D. Spatiotemporal traffic forecasting： review and proposed directions ［J］. Transport Reviews， 2018， 38（6）： 786-814.
[4]	Fu R， Zhang Z， Li L. Using LSTM and GRU neural network methods for traffic flow prediction ［C］// YAC 2016. Piscataway： IEEE， 2016： 324-328.
[5]	Zhao W， Gao Y， Ji T， et al. Deep temporal convolutional networks for short‑term traffic flow forecasting ［J］. IEEE Access， 2019， 7： 114496-114507.
[6]	Zheng C， Fan X， Wang C， et al. GMAN： a graph multi‑attention network for traffic prediction ［C］// AAAI 2020. Palo Alto： AAAI Press， 2020： 1234-1241.
[7]	Lv Z， Xu J， Zheng K， et al. LC‑RNN： a deep learning model for traffic speed prediction ［C］// IJCAI 2018. California： ijcai.org， 2018： 3470-3476.
[8]	Lu B， Gan X， Jin H， et al. Spatiotemporal adaptive gated graph convolution network for urban traffic flow forecasting ［C］// CIKM 2020. New York： ACM， 2020： 1025-1034.
[9]	Li Y， Yu R， Shahabi C， et al. Diffusion convolutional recurrent neural network： data‑driven traffic forecasting ［PP/OL］. V3. arXiv （2018-02-22）［2024-04-25］. .
[10]	Yu B， Yin H， Zhu Z. Spatio‑temporal graph convolutional networks： a deep learning framework for traffic forecasting ［C］// IJCAI 2018. California： ijcai.org， 2018： 3634-3640.
[11]	Wu Z， Pan S， Long G， et al. Graph WaveNet for deep spatial‑temporal graph modeling ［C］// IJCAI 2019. California： ijcai.org， 2019： 1907-1913.
[12]	Bai L， Yao L， Li C， et al. Adaptive graph convolutional recurrent network for traffic forecasting ［C］// NeurIPS 2020. Red Hook： Curran Associates Inc.， 2020： 17804-17815.
[13]	Chauhan N S， Kumar N， Eskandarian A. A novel confined attention mechanism driven Bi‑GRU model for traffic flow prediction ［J］. IEEE Transactions on Intelligent Transportation Systems， 2024， 25（8）： 9181-9191.
[14]	Chen J， Lv T， Cai S， et al. A novel detection model for abnormal network traffic based on bidirectional temporal convolutional network ［J］. Information and Software Technology， 2023， 157： No.107166.
[15]	Zhao L， Song Y， Zhang C， et al. T‑GCN： a temporal graph convolutional network for traffic prediction ［J］. IEEE Transactions on Intelligent Transportation Systems， 2020， 21（9）： 3848-3858.
[16]	Guo S， Lin Y， Feng N， et al. Attention based spatial‑temporal graph convolutional networks for traffic flow forecasting ［C］// AAAI 2019. Palo Alto： AAAI Press， 2019： 922-929.
[17]	Bai J， Zhu J， Song Y， et al. A3T‑GCN： attention temporal graph convolutional network for traffic forecasting ［J］. ISPRS International Journal of Geo‑Information， 2021， 10（7）： No.485.
[18]	Li F， Feng J， Yan H， et al. Dynamic graph convolutional recurrent network for traffic prediction： benchmark and solution ［J］. ACM Transactions on Knowledge Discovery from Data， 2023， 17（1）： No.9.
[19]	Lan S， Ma Y， Huang W， et al. DSTAGNN： dynamic spatial‑temporal aware graph neural network for traffic flow forecasting ［C］// ICML 2022. New York： JMLR.org， 2022： 11906-11917.
[20]	Sun Y， Jiang X， Hu Y， et al. Dual dynamic spatial‑temporal graph convolution network for traffic prediction ［J］. IEEE Transactions on Intelligent Transportation Systems， 2022， 23（12）： 23680-23693.
[21]	Chen C， Liu Y， Chen L， et al. Bidirectional spatial‑temporal adaptive transformer for urban traffic flow forecasting ［J］. IEEE Transactions on Neural Networks and Learning Systems， 2023， 34（10）： 6913-6925.
[22]	Ren Q， Li Y， Liu Y. Transformer‑enhanced periodic temporal convolution network for long short‑term traffic flow forecasting ［J］. Expert Systems with Applications， 2023， 227： No.120203.
[23]	Zhang Y， Liu S， Zhang P， et al. GRU‑ and Transformer‑based periodicity fusion network for traffic forecasting ［J］. Electronics， 2023， 12（24）： No.4988.
[24]	Yu Y， Xie X， Ju X. DPGARN： dynamic periodic graph attention recurrent network for traffic flow prediction ［C］// CCC 2024. Piscataway： IEEE， 2024： 8268-8273.
[25]	Lablack M， Yu S， Xu S， et al. Long‑sequence model for traffic forecasting in suboptimal situation ［C］// MobiArch 2023. New York： ACM， 2023： 25-30.
[26]	Mihaita A S， Li H， Rizoiu M A. Traffic congestion anomaly detection and prediction using deep learning ［PP/OL］. arXiv （2020-06-23）［2024-05-27］. .
[27]	Ye Y， Xiao Y， Zhou Y， et al. Dynamic multi‑graph neural network for traffic flow prediction incorporating traffic accidents ［J］. Expert Systems with Applications， 2023， 234： No.121101.
[28]	Chen J， Xu M， Xu W， et al. A flow feedback traffic prediction based on visual quantified features ［J］. IEEE Transactions on Intelligent Transportation Systems， 2023， 24（9）： 10067-10075.

数据集	节点数	时间片	边数	日期范围
PEMSD4	307	16 992	340	20180101/20180228
PEMSD8	170	17 856	277	20160701/20160831

数据集	节点数	时间片	边数	日期范围
PEMSD4	307	16 992	340	20180101/20180228
PEMSD8	170	17 856	277	20160701/20160831

数据集	模型	15 min			30 min			60 min			平均
数据集	模型	MAE	RMSE	MAPE/%	MAE	RMSE	MAPE/%	MAE	RMSE	MAPE/%	MAE	RMSE	MAPE/%
PEMSD4	DCRNN	20.61	32.22	13.91	23.87	36.80	16.18	30.83	46.27	21.27	24.44	37.48	16.64
	GWN	20.90	33.04	14.44	24.53	38.22	17.45	32.58	49.15	23.38	25.25	39.10	18.22
	AGCRN	18.88	30.92	12.49	19.65	32.29	12.94	21.34	34.98	13.87	19.76	32.48	12.99
	DDSTGCN	18.40	29.38	12.86	19.64	31.15	13.77	21.82	34.07	15.40	19.74	31.12	13.83
	Bi-STAT	18.16	29.29	12.39	19.01	30.61	12.92	21.68	33.85	15.36	19.66	30.64	12.76
	BDDGNet	17.44	28.82	11.70	18.25	30.31	12.25	19.54	32.27	13.10	18.26	30.26	12.16
PEMSD8	DCRNN	15.93	24.62	10.05	18.30	28.43	11.57	23.02	35.05	14.42	16.82	26.36	10.92
	GWN	15.59	25.27	9.84	18.28	30.02	12.06	24.18	38.85	15.63	18.28	30.05	12.15
	AGCRN	14.90	23.39	9.67	15.90	25.18	10.26	17.84	28.32	11.37	15.95	25.22	10.09
	DDSTGCN	14.31	22.45	9.71	15.44	24.42	10.39	17.42	27.33	11.61	15.49	24.36	10.42
	Bi-STAT	13.92	22.32	9.02	14.64	23.75	9.52	16.14	26.11	10.59	14.76	23.80	9.62
	BDDGNet	13.20	21.65	8.70	14.10	23.55	9.27	15.44	25.81	10.08	14.07	23.44	9.24

数据集	模型	15 min			30 min			60 min			平均
数据集	模型	MAE	RMSE	MAPE/%	MAE	RMSE	MAPE/%	MAE	RMSE	MAPE/%	MAE	RMSE	MAPE/%
PEMSD4	DCRNN	20.61	32.22	13.91	23.87	36.80	16.18	30.83	46.27	21.27	24.44	37.48	16.64
	GWN	20.90	33.04	14.44	24.53	38.22	17.45	32.58	49.15	23.38	25.25	39.10	18.22
	AGCRN	18.88	30.92	12.49	19.65	32.29	12.94	21.34	34.98	13.87	19.76	32.48	12.99
	DDSTGCN	18.40	29.38	12.86	19.64	31.15	13.77	21.82	34.07	15.40	19.74	31.12	13.83
	Bi-STAT	18.16	29.29	12.39	19.01	30.61	12.92	21.68	33.85	15.36	19.66	30.64	12.76
	BDDGNet	17.44	28.82	11.70	18.25	30.31	12.25	19.54	32.27	13.10	18.26	30.26	12.16
PEMSD8	DCRNN	15.93	24.62	10.05	18.30	28.43	11.57	23.02	35.05	14.42	16.82	26.36	10.92
	GWN	15.59	25.27	9.84	18.28	30.02	12.06	24.18	38.85	15.63	18.28	30.05	12.15
	AGCRN	14.90	23.39	9.67	15.90	25.18	10.26	17.84	28.32	11.37	15.95	25.22	10.09
	DDSTGCN	14.31	22.45	9.71	15.44	24.42	10.39	17.42	27.33	11.61	15.49	24.36	10.42
	Bi-STAT	13.92	22.32	9.02	14.64	23.75	9.52	16.14	26.11	10.59	14.76	23.80	9.62
	BDDGNet	13.20	21.65	8.70	14.10	23.55	9.27	15.44	25.81	10.08	14.07	23.44	9.24

数据集	头数	MAE	RMSE	MAPE/%
PEMSD4	1	18.32	30.64	12.33
	2	18.26	30.26	12.23
	7	18.35	30.46	12.40
PEMSD8	1	14.07	23.44	9.24
PEMSD8	3	14.20	23.64	9.30