A novel traffic accident prediction model was proposed, in order to reduce the occurrence of traffic accidents and property losses. The dual-scale decomposition equations are utilized to decompose the original traffic accident time series into a number of sub-layers, and the long short-term memory (LSTM) network is adopted to complete the forecasting of the low-frequency sub-layer. The double scale reconstruction equations are adopted to complete the predicted value reconstruction of the low-frequency sub-layer. LSTM, gate recurrent unit (GRU), stacked autoencoders (SAEs) and dual-scale LSTM (DS-LSTM) prediciton models were constructed, and the four models were used to predict the two data sets. Results show that compared with other models, the proposed model is robust and more effective in predicting the traffic accident time series. Compared with the original LSTM model, the prediction accuracy of model DS_LSTM is improved by 6% and 28% respectively in the two data sets. Testing on two different databases (Leeds and UK) shows that the proposed model has better generalization performance than the other models involved.
Wen-shu LI,Tao-tao ZOU,Hong-yan WANG,Hai HUANG. Traffic accident quantity prediction model based on dual-scale long short-term memory network. Journal of ZheJiang University (Engineering Science), 2020, 54(8): 1613-1619.
Fig.1Flow chart of traffic accident prediction model
Fig.2LSTM memory block structure
epochs
ACC/%
利兹数据集
UK数据集
400
74.8891
84.8406
500
79.3250
85.2503
600
79.9246
85.9217
700
81.5287
88.1876
800
82.0915
88.0292
900
83.2079
88.0172
1000
82.2952
85.4329
Tab.1Accuracy of DS-LSTM model under different epochs
Fig.3Accident number with epochs of 900 (Leeds database)
Fig.4Accident number with epochs of 700 (UK database)
模型
UK，epochs=700
Leeds，epochs=900
MAE
MSE
RMSE
ACC/%
MAE
MSE
RMSE
ACC/%
LSTM
0.5135
0.6161
0.7849
60.8588
0.4288
0.5811
0.7622
76.7608
GRU
0.6531
1.2837
1.1330
50.2231
1.0072
2.2285
1.4928
45.4190
SAEs
0.5872
0.8195
0.9052
55.2421
0.8611
1.4739
1.2140
53.3366
DS-LSTM
0.1900
0.0855
0.2924
88.1876
0.3227
0.2040
0.4517
83.2079
Tab.2Model characteristics under different data sets
[1]
ZIYAB A H, AKHTAR S Incidence and trend of road traffic injuries and related deaths in Kuwait: 2000–2009[J]. Injury, 2012, 43 (12): 2018- 2022
doi: 10.1016/j.injury.2011.09.023
[2]
YANNIS G, DRAGOMANOVITS A, LAIOU A, et al Road traffic accident prediction modelling: a literature review[J]. Proceedings of the Institution of Civil Engineers, 2017, 170 (5): 245- 254
[3]
ZHANG J, LIU X M, HE Y L, et al Application of ARIMA model in forecasting traffic accidents[J]. Journal of Beijing University of Technology, 2007, 33 (12): 1295- 1299
[4]
BARBA L, RODRIGUEZ N, MONTT C Smoothing strategies combined with ARIMA and neural networks to improve the forecasting of traffic accidents[J]. The Scientific World Journal, 2014, 1- 12
[5]
张艳艳, 刘晓佳, 熊子龙, 等 基于 ARIMA 模型的水上交通事故预测[J]. 中国水运: 下半月, 2017, (2): 51- 54 ZHANG Yan-yan, LIU Xiao-jia, XIONG Zi-long, et al Waterborne traffic accident prediction based on ARIMA model[J]. China Water Transport: the Second Half of the Month, 2017, (2): 51- 54
[6]
BECKER U, MANZ H Grey systems theory time series prediction applied to road traffic safety in Germany[J]. IFAC-PapersOnLine, 2016, 49 (3): 231- 236
doi: 10.1016/j.ifacol.2016.07.039
[7]
王星, 刘小勇 基于灰色马尔科夫模型的交通事故预测研究[J]. 交通科技与经济, 2017, 19 (4): 9- 13 WANG Xing, LIU Xiao-yong Study on traffic accident prediction based on grey Markov model[J]. Transportation Science and Technology, 2017, 19 (4): 9- 13
[8]
邸韡, 巴德瓦杰, 魏佳宁. 深度学习基础教程[M]. 北京: 机械工业出版社, 2018.
[9]
WANG J, GU Q, WU J, et al. Traffic speed prediction and congestion source exploration: a deep learning method [C]// IEEE 16th International Conference on Data Mining. Barcelona: IEEE, 2016: 499-508.
[10]
CONTRERAS E, TORRES-TREVINO L, TORRES F. Prediction of car accidents using a maximum sensitivity neural network [C]// 2018 EAI International Conference on Smart Technology. Monterrey: Springer, 2018: 86-95.
[11]
ZANG D, FANG Y, WANG D, et al. Long term traffic flow prediction using residual net and deconvolutional neural network [C]// Chinese Conference on Pattern Recognition and Computer Vision (PRCV). Guangzhou: Springer, 2018: 62-74.
[12]
GARCIADE SOTO B, BUMBACHER A, DEUBLEIN M, et al Predicting road traffic accidents using artificial neural network models[J]. Infrastructure Asset Management, 2018, 5 (4): 132- 144
doi: 10.1680/jinam.17.00028
[13]
ZHANG C, ZHANG H, YUAN D, et al Citywide cellular traffic prediction based on densely connected convolutional neural networks[J]. IEEE Communications Letters, 2018, 22 (8): 1656- 1659
doi: 10.1109/LCOMM.2018.2841832
[14]
PAMULA T Impact of data loss for prediction of traffic flow on an urban road using neural networks[J]. IEEE Transactions on Intelligent Transportation Systems, 2018, 20 (3): 1000- 1009
[15]
陈海龙, 彭伟 改进 BP 神经网络在交通事故预测中的研究[J]. 华东师范大学学报: 自然科学版, 2017, 2: 61- 68 CHEN Hai-long, PENG Wei Study on improved BP neural network in traffic accident prediction[J]. Journal of East China Normal University: Natural Science, 2017, 2: 61- 68
[16]
ZHAO Z, CHEN W, WU X, et al LSTM network: a deep learning approach for short-term traffic forecast[J]. IET Intelligent Transport Systems, 2017, 11 (2): 68- 75
doi: 10.1049/iet-its.2016.0208
[17]
CORTEZ B, CARRERA B, KIM Y J, et al An architecture for emergency event prediction using LSTM recurrent neural networks[J]. Expert Systems with Applications, 2018, 97: 315- 324
doi: 10.1016/j.eswa.2017.12.037
[18]
ALKHEDER S, TAAMNEH M, TAAMNEH S Severity prediction of traffic accident using an artificial neural network[J]. Journal of Forecasting, 2017, 36 (1): 100- 108
doi: 10.1002/for.2425
[19]
BEYLKIN G, COIFMAN R, ROKHLIN V Fast wavelet transforms and numerical algorithms I[J]. Communications on Pure and Applied Mathematics, 1991, 44 (2): 141- 183
doi: 10.1002/cpa.3160440202
[20]
MALLAT S G A theory for multiresolution signal decomposition: the wavelet representation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1989, (7): 674- 693
[21]
彭玉华. 小波变换与工程应用[M]. 北京: 科学出版社, 1999.
[22]
SHI X J, CHEN Z, WANG H, et al. Convolutional LSTM network: a machine learning approach for precipitation nowcasting [C]// Advances in Neural Information Processing Systems. 2015: 802-810.
[23]
RAHMANI B Early model of traffic sign reminder based on neural network[J]. Telkomnika, 2012, 10 (4): 749
doi: 10.12928/telkomnika.v10i4.864
[24]
CHAI T, DRAXLER R R Root mean square error (RMSE) or mean absolute error (MAE)? : arguments against avoiding RMSE in the literature[J]. Geoscientific Model Development, 2014, 7 (3): 1247- 1250
doi: 10.5194/gmd-7-1247-2014
[25]
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.
[26]
ZHANG D, KABUKA M R Combining weather condition data to predict traffic flow: a GRU-based deep learning approach[J]. IET Intelligent Transport Systems, 2018, 12 (7): 578- 585
doi: 10.1049/iet-its.2017.0313
