基于时空图卷积网络的交通事故风险预测研究

doi:10.19678/j.issn.1000-3428.0062961

摘要/Abstract

摘要： 交通事故的预测是通过对过去路段发生的交通事故进行分析，在综合考虑影响交通事故的相关因素后，对未来路段的交通事故发生状态进行预测。以往的大多数研究通常采用传统机器学习方法或单一深度学习模型预测法，利用网格化确定预测空间的单位，忽略了影响交通事故的天气、路况等外部因素，导致模型的预测性能不佳。提出一种基于时空特性的城市交通事故风险预测模型，在模型中使用改进的时空图卷积网络，利用图卷积网络（GCN）提取空间相关特征，并加入批标准化层解决梯度消失爆炸问题。在时间维度上采用门控线性单元（GLU）实现一维卷积操作，提取时间相关特征，并将GCN和GLU组合成时空卷积模块提取时空相关特征，使用均方误差损失函数解决样本数据零膨胀问题。实验结果表明，与GLU、SDCAE和ConvLSTM模型相比，该模型的RMSE指标分别降低了28%、4.87%、4.19%，能有效捕获时空相关性，综合性能得到较大提升。

关键词: 深度学习, 城市交通事故, 时空图卷积网络, 时空相关性, 批标准化层

Abstract: Traffic accident prediction is the analysis of past road traffic accident data and other related factors to predict the future state of road traffic accidents.Most of the previous research methods have typically been traditional machine-learning methods or a single deep-learning model prediction method utilizing a grid to determine the prediction space unit, weather, road conditions, and other external factors affecting traffic accidents.However, these methods ignore the problem of zero expansion, which leads to poor prediction performance.Therefore, this paper presents a prediction model for urban traffic accident risk considering spatio-temporal characteristics.The Improved Spatio-Temporal Graph Convolution Network (ISTGCN) is used in the model.First, a Graph Convolution Network (GCN) is used to extract spatial correlation features, and Batch Normalization (BN) layer is added to solve the gradient-disappearing explosion problem.Second, Gated Linear Units (GLU) are used to extract time-dependent features by applying a one-dimensional convolution operation.Finally, the GCN and GLU are combined into a Space-Time Convolution (ST-Conv) module to extract the spatio-temporal correlation features and the Mean Squared Error(MSE) loss function is used to solve the problem of zero expansion of sample data.Experimental results show that compared with the GLU, SDCAE, and ConvLSTM models, the Root Mean Square Error (RMSE) of this model is reduced by 28%, 4.87%, and 4.19% respectively.Thus, this method can effectively capture the spatio-temporal correlation and improve the comprehensive performance.

Key words: deep learning, urban traffic accident, Spatio-Temporal Graph Convolution (STGCN), spatio-temproal correlation, Batch Normalization(BN) layer

中图分类号:

TP393

王庆荣, 魏怡萌, 朱昌锋, 田可可. 基于时空图卷积网络的交通事故风险预测研究[J]. 计算机工程, 2022, 48(11): 22-29.

WANG Qingrong, WEI Yimeng, ZHU Changfeng, TIAN Keke. Research on Traffic Accident Risk Prediction Based on Spatio-Temporal Graph Convolutional Network[J]. Computer Engineering, 2022, 48(11): 22-29.

http://www.ecice06.com/CN/Y2022/V48/I11/22

图/表 12

20230203202143

20230203202147

20230203202150

20230203202154

20230203202157

20230203202200

20230203202204

20230203202207

20230203202210

20230203202214

20230203202217

20230203202221

参考文献

[1] FITZPATRICK L.Global status report on road safety 2015:time for action[EB/OL].[2021-09-04].https://www.abe.pl/pl/book/9789241565066/global-status-report-on-road-safety-2015-time-for-action.
[2] ABDEL-ATY M, UDDIN N, PANDE A, et al.Predicting freeway crashes from loop detector data by matched case-control logistic regression[J].Journal of the Transportation Research Board, 2004, 1897(1):88-95.
[3] TIAN Z, ZHANG S R.Application of big data optimized clustering algorithm in cloud computing environment in traffic accident forecast[J].Peer-to-Peer Networking and Applications, 2021, 14(4):2511-2523.
[4] MA X B, ZHANG J Y, HUANG P, et al.Traffic accident prediction based on Markov chain cloud model[EB/OL].[2021-09-04].https://www.researchgate.net/publication/342785308_Traffic_accident_prediction_based_on_Markov_chain_cloud_model.
[5] ROLAND J, WAY P D, FIRAT C, et al.Modeling and predicting vehicle accident occurrence in Chattanooga, Tennessee[J].Accident Analysis & Prevention, 2021, 149:60-68.
[6] 邓晓庆, 孟祥海, 郑来.基于BP神经网络的高速公路事故预测模型[J].交通信息与安全, 2016, 34(1):78-84. DENG X Q, MENG X H, ZHENG L.An accident prediction model for expressway based on BP neural network[J].Journal of Transport Information and Safety, 2016, 34(1):78-84.(in Chinese)
[7] 宋英华, 程灵希, 刘丹, 等.基于组合预测优化模型的交通事故预测研究[J].中国安全科学学报, 2017, 27(5):31-35. SONG Y H, CHENG L X, LIU D, et al.Research on optimal combination model for forecasting traffic accident[J].China Safety Science Journal, 2017, 27(5):31-35.(in Chinese)
[8] 胡立伟, 张婷, 郭凤香, 等.基于灰色BP神经网络的道路交通事故车型分担率预测及其预防策略研究[J].武汉理工大学学报(交通科学与工程版), 2018, 42(3):388-392, 397. HU L W, ZHANG T, GUO F X, et al.Traffic accident split rate of vehicle types prediction and prevention strategies study based on gray BP neural network[J].Journal of Wuhan University of Technology (Transportation Science & Engineering), 2018, 42(3):388-392, 397.(in Chinese)
[9] 李志慧, 孙雅倩, 陶鹏飞, 等.交通事故后的交通运行风险状态等级预测方法[J].吉林大学学报(工学版), 2021, 52(1):127-135. LI Z H, SUN Y Q, TAO P F, et al.Prediction method of traffic operation risk level after traffic accident[J].Journal of Jilin University(Engineering and Technology Edition), 2021, 52(1):127-135.(in Chinese)
[10] CHEN Q J, SONG X, YAMADA H, et al.Learning deep representation from big and heterogeneous data for traffic accident inference[C]//Proceedings of the 30th Conference on Artificial Intelligence.[S.l.]:AAAI Press, 2016:338-344.
[11] 李文书, 邹涛涛, 王洪雁, 等.基于双尺度长短期记忆网络的交通事故量预测模型[J].浙江大学学报(工学版), 2020, 54(8):1613-1619. LI W S, ZOU T T, WANG H Y, et al.Traffic accident quantity prediction model based on dual-scale long short-term memory network[J].Journal of Zhejiang University (Engineering Science), 2020, 54(8):1613-1619.(in Chinese)
[12] CHEN C, FAN X L, ZHENG C P, et al.SDCAE:stack denoising convolutional autoencoder model for accident risk prediction via traffic big data[C]//Proceedings of the 6th International Conference on Advanced Cloud and Big Data.Washington D.C., USA:IEEE Press, 2018:328-333.
[13] ZHU L, LI T R, DU S D.TA-STAN:a deep spatial-temporal attention learning framework for regional traffic accident risk prediction[C]//Proceedings of International Joint Conference on Neural Networks.Washington D.C., USA:IEEE Press, 2019:1-8.
[14] 赵海涛, 程慧玲, 丁仪, 等.基于深度学习的车联边缘网络交通事故风险预测算法研究[J].电子与信息学报, 2020, 42(1):50-57. ZHAO H T, CHENG H L, DING Y, et al.Research on traffic accident risk prediction algorithm of edge Internet of vehicles based on deep learning[J].Journal of Electronics & Information Technology, 2020, 42(1):50-57.(in Chinese)
[15] 张延孔, 卢家品, 张帅超, 等.基于路网结构的城市交通事故短期风险预测方法[J].智能系统学报, 2020, 15(4):663-671. ZHANG Y K, LU J P, ZHANG S C, et al.A short-term risk prediction method for urban traffic accidents based on road network[J].CAAI Transactions on Intelligent Systems, 2020, 15(4):663-671.(in Chinese)
[16] GORI M, MONFARDINI G, SCARSELLI F.A new model for learning in graph domains[C]//Proceedings of 2005 IEEE International Joint Conference on Neural Networks.Washington D.C., USA:IEEE Press, 2005:729-734.
[17] 荣斌, 武志昊, 刘晓辉, 等.基于时空多图卷积网络的交通站点流量预测[J].计算机工程, 2020, 46(5):26-33. RONG B, WU Z H, LIU X H, et al.Flow prediction of traffic stations based on spatio-temporal multi-graph convolutional network[J].Computer Engineering, 2020, 46(5):26-33.(in Chinese)
[18] 李社蕾, 周波, 杨博雄, 等.图卷积神经网络基础图结构影响研究[J].小型微型计算机系统, 2021, 42(4):891-896. LI S L, ZHOU B, YANG B X, et al.Study on the influence of basic graph structure in graph convolutional network[J].Journal of Chinese Computer Systems, 2021, 42(4):891-896.(in Chinese)
[19] MASON J C, HANDSCOMB D C.Chebyshev polynomials[EB/OL].[2021-09-04].https://www.scirp.org/reference/ReferencesPapers.aspx?ReferenceID=575225.
[20] YU B, YIN H T, ZHU Z X.Spatio-temporal graph convolutional networks:a deep learning framework for traffic forecasting[EB/OL].[2021-09-04].https://arxiv.org/abs/1709.04875.
[21] 李俊辉, 汤左淦.交通事故伤害程度估计中的零膨胀现象研究[J].公路工程, 2020, 45(5):239-245. LI J H, TANG Z G.An analysis for addressing excess zeros phenomenon in estimating traffic crash injury severity[J].Highway Engineering, 2020, 45(5):239-245.(in Chinese)
[22] 王彦浩.基于图神经网络的交通事故预测方法研究[D].北京:华北电力大学, 2021. WANG Y H.Research on traffic accident prediction based on graph neural network[D].Beijing:North China Electric Power University, 2021.(in Chinese)
[23] CHEN T Q, GUESTRIN C.XGBoost:a scalable tree boosting system[C]//Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.New York, USA:ACM Press, 2016:785-794.
[24] CHUNG J, GULCEHRE C, CHO K, et al.Empirical evaluation of gated recurrent neural networks on sequence modeling[EB/OL].[2021-09-04].https://arxiv.org/abs/1412.3555.
[25] SHI X J, CHEN Z R, WANG H, et al.Convolutional LSTM network:a machine learning approach for precipitation nowcasting[EB/OL].[2021-09-04].https://arxiv.org/abs/1506.04214.

选择文件类型/文献管理软件名称

选择包含的内容