Please wait a minute...
Journal of ZheJiang University (Engineering Science)  2021, Vol. 55 Issue (10): 1825-1833    DOI: 10.3785/j.issn.1008-973X.2021.10.004
    
Adaptive graph generation jump network for traffic flow prediction
Jing HUANG1(),Shu-yuan ZHONG1,Yuan-qiao WEN2,Kun LUO1
1. School of Computer Science and Technology, Wuhan University of Technology, Wuhan 430063, China
2. Intelligent Transportation System Research Center, Wuhan University of Technology, Wuhan 430063, China
Download: HTML     PDF(1071KB) HTML
Export: BibTeX | EndNote (RIS)      

Abstract  

A novel deep-learning-based model, adaptive graph generation jump network (AG-JNet), was proposed to solve the problem that traffic flow data has complex spatial-temporal correlations. The model consisted of two spatial-temporal modules, each of which was divided into two critical components, i.e., temporal correlation block and spatial correlation block. The temporal correlation block used multi-layer dilated convolution to increase the receptive field in temporal dimension while reducing computational cost. The spatial correlation block used adaptive graph generation convolution, which did not rely on the fixed graph structure to extract spatial correlation. Stacking multiple layers by jumping connections was used in both temporal and spatial modeling in order to improve the ability of extracting deep features of the model. The temporal feature and the spatial feature were fused by gated mechanism to obtain the spatial-temporal features for traffic flow prediction. Extensive experiments were conducted on two public datasets, i.e., PeMSD4 and PeMSD8. The experimental results showed that the AG-JNet achieved excellent performance under different traffic indicators.



Key wordstraffic flow prediction      spatial-temporal correlation      adaptive graph generation      dilated convolution      jumping connection     
Received: 10 December 2020      Published: 27 October 2021
CLC:  TP 399  
Fund:  国家自然科学基金资助项目(52072287);武汉理工大学自主创新研究基金资助项目(205210016)
Cite this article:

Jing HUANG,Shu-yuan ZHONG,Yuan-qiao WEN,Kun LUO. Adaptive graph generation jump network for traffic flow prediction. Journal of ZheJiang University (Engineering Science), 2021, 55(10): 1825-1833.

URL:

https://www.zjujournals.com/eng/10.3785/j.issn.1008-973X.2021.10.004     OR     https://www.zjujournals.com/eng/Y2021/V55/I10/1825


用于交通流预测的自适应图生成跳跃网络

针对交通流数据复杂的时空相关性,提出新的基于深度学习的自适应图生成跳跃网络(AG-JNet模型). 该模型由2个时空模块组成,每个时空模块分为2支,分别对时间相关性和空间相关性建模. 时间建模采用多层扩张卷积,在增大时间维度感受野的同时降低计算开销. 空间建模采用自适应图生成卷积,在不依赖图的固定结构下提取空间相关性. 在时间和空间的建模中均采用跳跃连接堆叠多层,以提升模型的深层特征提取能力,将时间特征和空间特征进行门控融合,提取出用于交通流量预测的时空特征. 在2个真实数据集PeMSD4和PeMSD8上的实验表明,AG-JNet在不同指标下取得了优异的性能.


关键词: 交通流量预测,  时空相关性,  自适应图生成,  扩张卷积,  跳跃连接 
Fig.1 Overview of adaptive graph generation jump networks
Fig.2 Illustration of spatial-temporal blocks
Fig.3 Two-layer graph convolution
Fig.4 Dilated convolution in temporal dimension
Fig.5 Gated fusion
模型 PeMSD4 (30/60 min) PeMSD8 (30/60 min)
MAE RMSE MAE RMSE
HA 31.46/38.24 44.94/54.31 26.04/31.94 36.99/45.16
T-GCN 22.97/24.8 34.74/37.09 20.17/21.52 29.03/31.19
STGCN 22.85/25.8 33.62/37.68 19.85/22.27 28.25/31.52
ASTGCN 21.69/23.14 33.72/36.11 17.67/19.11 27.59/30.27
AGCRN 18.72/19.74 30.65/32.57 15.00/16.58 23.69/26.19
AG-JNet 18.53/20.08 28.19/30.43 15.15/16.12 22.80/24.59
Tab.1 Performance comparison of different methods on two data sets
Fig.6 Performance comparison of related methods under different forecasting time steps
模型 参数规模 ttr /s tt /s RMSE
T-GCN 13452 10.53 0.32 31.19
STGCN 99124 6.45 0.49 31.52
ASTGCN 560604 47.19 7.41 30.27
AGCRN 150112 24.15 3.32 26.19
AG-JNet 241079 18.27 1.84 24.59
Tab.2 Time-consuming comparison of different traffic flow prediction models
Fig.7 Results of ablation experiments
Fig.8 Comparison of smoothness metric values
[1]   WILLIAMS B M, HOEL L A Modeling and forecasting vehicular traffic flow as a seasonal ARIMA process: theoretical basis and empirical results[J]. Journal of Transportation Engineering, 2003, 129 (6): 664- 672
doi: 10.1061/(ASCE)0733-947X(2003)129:6(664)
[2]   HAMED M M, AL-MASAEID H R, SAID Z M B Short-term prediction of traffic volume in urban arterials[J]. Journal of Transportation Engineering, 1995, 121 (3): 249- 254
doi: 10.1061/(ASCE)0733-947X(1995)121:3(249)
[3]   ZIVOT E, WANG J. Vector autoregressive models for multivariate time series [M]// Modeling financial time series with S-PLUS®. New York: Springer, 2006: 385-429.
[4]   FU R, ZHANG Z, LI L. Using LSTM and GRU neural network methods for traffic flow prediction [C]// 2016 31st Youth Academic Annual Conference of Chinese Association of Automation. Wuhan: IEEE, 2016: 324-328.
[5]   LI Y, YU R, SHAHABI C, et al. Diffusion convolutional recurrent neural network: data-driven traffic forecasting [EB/OL]. [2020-10-10]. https://arxiv.org/pdf/1707.01926.pdf.
[6]   YU B, YIN H, ZHU Z. Spatio-temporal graph convolutional networks: a deep learning framework for traffic forecasting [EB/OL]. [2020-10-10]. https://arxiv.org/pdf/1709.04875.pdf.
[7]   GUO S, LIN Y, FENG N, et al. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting [C]// Proceedings of the AAAI Conference on Artificial Intelligence. Hawaii: AAAI, 2019: 922-929.
[8]   BAI L, YAO L, KANHERE S S, et al. Passenger demand forecasting with multi-task convolutional recurrent neural networks[C]// Pacific-Asia Conference on Knowledge Discovery and Data Mining. Cham: Springer, 2019: 29-42.
[9]   GENG X, LI Y, WANG L, et al. Spatiotemporal multi-graph convolution network for ride-hailing demand forecasting [C]// Proceedings of the AAAI Conference on Artificial Intelligence. Hawaii: AAAI, 2019: 3656-3663.
[10]   KIPF T N, WELLING M. Semi-supervised classification with graph convolutional networks [EB/OL]. [2020-10-10]. https://arxiv.org/pdf/1609.02907.pdf.
[11]   XU K, LI C, TIAN Y, et al. Representation learning on graphs with jumping knowledge networks [EB/OL]. [2020-10-10]. https://arxiv.org/pdf/1806.03536.pdf.
[12]   AHMED M S, COOK A R Analysis of freeway traffic time-series data by using Box-Jenkins techniques[J]. Transportation Research Record Journal of the Transportation Research Board, 1979, 773 (722): 1- 9
[13]   PARK D, RILETT L R Forecasting freeway link travel times with a multilayer feedforward neural network[J]. Computer-Aided Civil and Infrastructure Engineering, 1999, 14 (5): 357- 367
doi: 10.1111/0885-9507.00154
[14]   HUANG W, SONG G, HONG H, et al Deep architecture for traffic flow prediction: deep belief networks with multitask learning[J]. IEEE Transactions on Intelligent Transportation Systems, 2014, 15 (5): 2191- 2201
doi: 10.1109/TITS.2014.2311123
[15]   WU Y, TAN H. Short-term traffic flow forecasting with spatial-temporal correlation in a hybrid deep learning framework [EB/OL]. [2020-10-10]. https://arxiv.org/pdf/1612.01022.pdf.
[16]   YU H, WU Z, WANG S, et al Spatiotemporal recurrent convolutional networks for traffic prediction in transportation networks[J]. Sensors, 2017, 17 (7): 1501
doi: 10.3390/s17071501
[17]   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, 2019, 21 (9): 3848- 3858
[18]   冯宁, 郭晟楠, 宋超, 等 面向交通流量预测的多组件时空图卷积网络[J]. 软件学报, 2019, 30 (3): 759- 769
FENG Ning, GOU Sheng-nan, SONG Chao, et al Multi-component spatial-temporal graph convolution networks for traffic flow forecasting[J]. Journal of Software, 2019, 30 (3): 759- 769
[19]   ZHENG C, FAN X, WANG C, et al. Gman: a graph multi-attention network for traffic prediction [C]// Proceedings of the AAAI Conference on Artificial Intelligence. New York: AAAI, 2020: 1234-1241.
[20]   CHEN W, CHEN L, XIE Y, et al. Multi-range attentive bicomponent graph convolutional network for traffic forecasting [EB/OL]. [2020-10-10]. https://arxiv.org/ftp/arxiv/papers/1911/1911.12093.pdf.
[21]   FANG S, ZHANG Q, MENG G, et al. GSTNet: Global spatial-temporal network for traffic flow prediction [C]// 28th International Joint Conference on Artificial Intelligence. Macao: Morgan Kaufmann, 2019: 2286-2293.
[22]   BAI L, YAO L, LI C, et al. Adaptive graph convolutional recurrent network for traffic forecasting [EB/OL]. [2020-10-10]. https://arxiv.org/pdf/2007.02842.pdf.
[23]   HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
[24]   IOFFE S, SZEGEDY C. Batch normalization: accelerating deep network training by reducing internal covariate shift [EB/OL]. [2020-10-10]. https://arxiv.org/pdf/1502.03167.pdf.
[25]   HU J, SHEN L, SUN G. Squeeze-and-excitation networks [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 7132-7141.
[26]   HE K, ZHANG X, REN S, et al. Delving deep into rectifiers: surpassing human-level performance on ImageNet classification [C]// Proceedings of the IEEE International Conference on Computer Vision. Santiago: IEEE, 2015: 1026-1034.
[1] GONG Yue, LUO Xiao-Qin, WANG Dian-hai, YANG Shao-hui. Urban travel time prediction based on gradient boosting regression tress[J]. Journal of ZheJiang University (Engineering Science), 2018, 52(3): 453-460.