Traffic Flow Prediction with Missing Data Based on Spatial-Temporal Convolutional Neural Networks

doi:10.3778/j.issn.1002-8331.2108-0329

Abstract

Abstract: In order to improve the accuracy of traffic flow prediction under the condition of continuous missing data, a spatial-temporal convolutional neural network prediction algorithm is proposed. The spatial-temporal matrix of road network traffic data is established by considering the spatial adjacent relationship of horizontal distribution and the time dependence of vertical distribution. A mask matrix is introduced to represent the missing situation of data. The convolution operation is used to mine the implicit nonlinear correlation between the upstream and downstream detectors, establish the mapping relationship between the current and future traffic states, and realize the traffic flow prediction under the condition of continuous data absence. The experimental results show that the accuracy of traffic flow prediction is better than those of LSTM, GRU and GMN models in the absence of continuous traffic data. The model shows good stability and robustness, and improves the accuracy of traffic flow prediction in the absence of data.

Key words: intelligent transportation system, traffic flow forecasting, deep learning, spatial-temporal convolutional neural network, continuous data missing

摘要： 针对数据连续缺失情况下交通流预测精度下降甚至失效的问题，提出了一种时空卷积神经网络预测模型，根据横向分布的时间相关性和纵向分布的空间相关性，构建路网交通数据时空矩阵，引入掩码矩阵标记数据的缺失状况，利用卷积操作提取路网中各检测器之间隐含的非线性关联，建立当前时刻与未来交通状态的映射关系，实现数据缺失情况下的交通流预测。使用公开数据集，在3个时间尺度上的验证结果表明，所提出的模型在平均误差和预测精度两个方面均优于长短期记忆网络、门控循环单元、扩散卷积神经网络和图马尔可夫网络模型，在交通数据随机缺失和连续缺失两种情况下，均表现出了良好的稳定性和健壮性。

关键词: 智能交通系统, 交通流预测, 深度学习, 时空卷积神经网络, 连续数据缺失

ZHANG Zhuangzhuang, QU Licheng, LI Xiang, ZHANG Minghao, LI Zhaolu. Traffic Flow Prediction with Missing Data Based on Spatial-Temporal Convolutional Neural Networks[J]. Computer Engineering and Applications, 2022, 58(7): 259-265.

张壮壮, 屈立成, 李翔, 张明皓, 李昭璐. 基于时空卷积神经网络的数据缺失交通流预测[J]. 计算机工程与应用, 2022, 58(7): 259-265.

References

[1] QU L C，LV J，LI W，et al.Features injected recurrent neural networks for short-term traffic speed prediction[J].Neurocomputing，2021，451：290-304.
[2] 谷振宇，陈聪，郑家佳，等.基于时空图卷积循环神经网络的交通流预测[J/OL].控制与决策：1-9[2021-06-24].https：//doi.org/10.13195/j.kzyjc.2020.1445.
GU Z Y，CHEN C，ZHENG J J，et al.Based on the space-time map convolution cycle neural network of traffic flow prediction[J/OL].Control and Decision：1-9[2021-06-24].https：//doi.org/10.13195/j.kzyjc.2020.1445.
[3] 康梦轩，宋俊平，范鹏飞，等.基于深度学习的网络流量预测研究综述[J].计算机工程与应用，2021，57（10）：1-9.
KANG M X，SONG J P，FAN P F，et al.Survey of network traffic forecast based on deep learning[J].Computer Engineering and Applications，201，57（10）：1-9.
[4] 刘明宇，吴建平，王钰博，等.基于深度学习的交通流量预测[J].系统仿真学报，2018，30（11）：4100-4105.
LIU M Y，WU J P，WANG Y B，et al.Traffic flow prediction based on deep learning[J].Journal of System Simulation，2018，30（11）：4100-4105.
[5] 戢晓峰，戈艺澄.基于深度学习的节假日高速公路交通流预测方法[J].系统仿真学报，2020，32（6）：1164-1171.
JI X F，GE Y C.Holiday highway traffic flow prediction method based on deep learning[J].Journal of System Simulation，2020，32（6）：1164-1171.
[6] LI Y，YU R，SHAHABI C，et al.Diffusion convolutional recurrent neural network：Data-driven traffic forecasting[C]//Proceedings of International Conference on Learning Representations，2018.
[7] 刘钊，杜威，闫冬梅，等.基于K近邻算法和支持向量回归组合的短时交通流预测[J].公路交通科技，2017，34（5）：122-128.
LIU Z，DU W，YAN D M，et al.Short-term traffic flow prediction based on K-nearest neighbor algorithm and support vector regression combination[J].Highway Traffic Science and Technology，2017，34（5）：122-128.
[8] 曹洁，尹阳阳，张红，等.改进FCM的交通流缺失数据修复方法[J].传感器与微系统，2020，39（10）：54-57.
CAO J，YIN Y Y，ZHANG H，et al.Improved FCM for missing data repair of traffic flow[J].Sensor and MicroSystem，2020，39（10）：54-57.
[9] CHEN X B，CAI Y F，YE Q L，et al.Graph regularized local self?representation for missing value imputation with applications to on-road traffic sensor data[J].Neurocomputing，2018，303：47-59.
[10] 马宝泽，张天骐，安泽亮，等.基于张量分解的卷积盲源分离方法[J].通信学报，2021，42（8）：52-60.
MAB Z，ZHANG T Q，AN Z L，et al.Convolutional blind source separation method based on tensor decomposition[J].Journal of Communication，2021，42（8）：52-60.
[11] CHEN X Y，HE Z C，SUN L S.A Bayesian tensor decomposition approach for spatiotemporal traffic data imputation[J].Transportation Research Part C Emerging Technologies，2018，98：73-84.
[12] 何领朝，林东，冯心欣.基于自适应秩动态张量分析的短时交通流预测[J].物联网学报，2019，3（3）：18-25.
HE L C，LIN D，FENG X X.Short?term traffic flow prediction based on adaptive rank dynamic tensor analysis[J].Journal of the Internet of Things，2019，3（3）：18-25.
[13] CUI Z，LIN L，PU Z，et al.Graph Markov network for traffic forecasting with missing data[J].Transportation Research Part C Emerging Technologies，2020，117：102671.
[14] WHITLOCK M E，QUEEN C M.Modelling a traffic network with missing data[J].Journal of Forecasting，2000，19（7）：561-574.
[15] 刘博，王明烁，李永，等.深度学习在时空序列预测中的应用综述[J].北京工业大学学报，2021，47（8）：925-941.
LIU B，WANG M S，LI Y，et al.A review of applications of deep learning in spatio-temporal series prediction[J].Journal of Beijing University of Technology，2021，47（8）：925-941.