Graph Neural Network for Traffic Flow Situation Prediction

doi:10.3778/j.issn.1673-9418.2011006

Abstract

Abstract:

Road network structure integrated traffic flow situation prediction is a highly nonlinear and complexly spatial-temporal dynamic correlation time-series data prediction problem. However, traditional traffic flow situation forecasting methods cannot model the temporal and spatial correlation in long-term series data in the traffic network. To address the issue mentioned above, a deep learning model of traffic flow prediction based on graph structure is proposed. Firstly, the graph wavelet convolution operator is defined based on the graph wavelet transform. Furthermore, the graph wavelet convolution neural network module is designed based on the operator for the traffic flow situation prediction. Secondly, a spatial-temporal dynamic correlation model is constructed based on the spatial-temporal attention mechanism to capture the dynamic temporal and spatial correlation of the traffic network. Finally, the strategy of stacking multi-layer graph wavelet neural network modules is adopted to establish a novel graph wavelet neural network for road network traffic flow situation prediction. Experimental results show that the developed model??s performance on the experimental datasets is better than the existing baseline models. In the comparative experiment on the non-zero element statistics of the graph wavelet transform matrix and the Fourier transform matrix, it??s found that the convolution operation based on the graph wavelet transform is more sparse. Therefore, the convolution operation defined based on the graph wavelet transform is more helpful to improve the calculation efficiency of the traffic flow situation prediction model.

Key words: traffic flow situation prediction, graph convolution, graph wavelet neural network, traffic flow

摘要：

融合了路网结构的交通流态势预测是一个高度非线性化且复杂的时空动态相关性的时序数据预测问题。然而，传统交通流态势预测方法无法建模交通网络中长时间序列数据间的时空相关性。针对交通路网交通流态势预测问题，提出了一种基于图结构的交通流预测深度学习模型。首先，基于图小波变换定义图小波卷积算子，设计了面向路网交通流态势预测的图小波卷积神经网络模块；其次，结合时空注意机制构建了用于道路网络交通流态势预测的时空动态相关性模型，以捕获交通网络的动态时空相关性；最后，采用叠加多层图小波神经网络模块的策略，构建了一种面向路网交通流态势预测的图小波卷积神经网络模型。实验结果表明，该网络模型在数据集上的性能优于现有的基线模型。通过图小波变换矩阵与傅里叶变换矩阵非零元素统计对比实验，发现基于图小波变换定义的卷积运算更具稀疏性。因此，基于图小波变换定义的卷积运算更有助于提升交通流态势预测模型的计算效率。

关键词: 交通流态势预测, 图卷积, 图小波神经网络, 交通流

JIANG Shan, DING Zhiming, XU Xinrun, YAN Jin. Graph Neural Network for Traffic Flow Situation Prediction[J]. Journal of Frontiers of Computer Science and Technology, 2021, 15(6): 1084-1091.

姜山, 丁治明, 徐馨润, 严瑾. 面向路网交通流态势预测的图神经网络模型[J]. 计算机科学与探索, 2021, 15(6): 1084-1091.

References

[1] QIN W J, JIA W, HAN F, et al. Superpixel-based and boundary-sensitive convolutional neural network for automated liver segmentation[J]. Physics in Medicine & Biology, 2018, 63(9): 095017.
[2] ROCCO I, ARANDJELOVIC R, SIVIC J. Convolutional neural network architecture for geometric matching[J]. IEEE Trans-actions on Pattern Analysis and Machine Intelligence, 2019, 41(11): 2553-2567.
[3] LV Y S, DUAN Y J, KANG W W, et al. Traffic flow predi-ction with big data: a deep learning approach[J]. IEEE Trans-actions on Intelligent Transportation Systems, 2015, 16(2): 865-873.
[4] FOULADGAR M, PARCHAMI M, ELMASRI R, et al. Sca-lable deep traffic flow neural networks for urban traffic con-gestion prediction[C]//Proceedings of the 2017 International Joint Conference on Neural Networks, Anchorage, May 14-19, 2017. Piscataway: IEEE, 2017: 2251-2258.
[5] PENG S F, SHEN Y Y, ZHU Y M, et al. A frequency-aware spatio-temporal network for traffic flow prediction[C]//LNCS 11447: Proceedings of the 24th International Conference on Database Systems for Advanced Applications, Chiang Mai, Apr 22-25, 2019. Berlin, Heidelberg: Springer, 2019: 697-712.
[6] SONG T F, ZHENG W M, SONG P, et al. EEG emotion reco-gnition using dynamical graph convolutional neural networks[J]. IEEE Transactions on Affective Computing, 2020, 11(3): 532-541.
[7] XU Y, KONG Q Q, HUANG Q, et al. Convolutional gated recurrent neural network incorporating spatial features for audio tagging[C]//Proceedings of the 2017 International Joint Conference on Neural Networks, Anchorage, May 14-19, 2017. Piscataway: IEEE, 2017: 3461-3466.
[8] SOUMYASUNDAR P, FLORENCE R, MARK C. Bayesian graph convolutional neural networks using node copying[C]//Proceedings of the International Conference on Machine Learning, Learning and Reasoning with Graph-Structured Representations Workshop, Long Beach, Jun 6, 2019: 1-6.
[9] YOUNGJOO S, MICHAEL D, PIERRE V, et al. Structured sequence modeling with graph convolutional recurrent networks[J]. arXiv:1612.07659, 2016.
[10] GREGOR K, DANIHELKA I, GRAVES A, et al. DRAW: a recurrent neural network for image generation[C]//Proceed-ings of the 32nd International Conference on Machine Lear-ning, Lille, Jul 6-11, 2015: 1462-1471.
[11] YU B, YIN H T, ZHU Z X. Spatio-temporal graph convolu-tional networks: a deep learning framework for traffic fore-casting[C]//Proceedings of the 27th International Joint Con-ference on Artificial Intelligence, Stockholm, Jul 13-19, 2018: 3634-3640.
[12] WU Z H, PAN S R, LONG G D, et al. Graph WaveNet for deep spatial-temporal graph modeling[C]//Proceedings of the 28th International Joint Conference on Artificial Intelligence, Macao, China, Aug 10-16, 2019: 1907-1913.
[13] XU B B, SHEN H W, CAO Q, et al. Graph wavelet neural network[C]//Proceedings of the 7th International Conference on Learning Representations, New Orleans, May 6-9, 2019: 1-14.
[14] DONNAT C, ZITNIK M, HALLAC D, et al. Learning structural node embeddings via diffusion wavelets[C]//Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, London, Aug 19-23, 2018. New York: ACM, 2018: 1320-1329.
[15] HAMMOND D K, VANDERGHEYNST P, GRIBONVAL R. Wavelets on graphs via spectral graph theory[J]. Applied and Computational Harmonic Analysis, 2011, 30: 129-150.
[16] TREMBLAY N, BORGNAT P. Graph wavelets or multiscale community mining[J]. IEEE Transactions on Signal Proces-sing, 2014, 62(20): 5227-5239.
[17] SUN X Q, FENG Y J. Sensitivity analysis of multilayer perceptron[J]. Chinese Journal of Computers, 2001, 24(9): 951-958.
孙学全, 冯英浚. 多层感知器的灵敏度分析[J]. 计算机学报, 2001, 24(9): 951-958.
[18] LIANG Y X, KE S, ZHANG J B, et al. GeoMAN: multi-level attention networks for geo-sensory time series predic-tion[C]//Proceedings of the 27th International Joint Conference on Artificial Intelligence, Stockholm, Jul 13-19, 2018: 3428-3434.
[19] CHEN C, PETTY K F, SKABARDONIS A, et al. Freeway performance measurement system: mining loop detector data[J]. Transportation Research Record: Journal of the Transportation Research Board, 2001(1): 96-102.
[20] GUO S N, LI Y F, FENG N, et al. Attention based spatial-temporal graph convolutional networks for traffic flow fore-casting[C]//Proceedings of the 33rd AAAI Conference on Artificial Intelligence, the 31st Innovative Applications of Artificial Intelligence Conference, the 9th AAAI Symposium on Educational Advances in Artificial Intelligence, Honolulu, Jan 27-Feb 1, 2019. Menlo Park: AAAI, 2019: 922-929.
[21] ZIVOT E, WANG J H. Vector autoregressive models for multivariate time series[M]//ZIVOT E, WANG J H. Modeling Financial Time Series with S-Plus?. Berlin, Heidelberg: Springer, 2003.