交通事故后的交通运行风险状态等级预测方法

doi:10.13229/j.cnki.jdxbgxb20200731

摘要/Abstract

摘要：

以事故数据为基础研究事后风险状态，建立了基于改进深度森林算法的交通运行风险状态等级预测模型。首先分析了事故特征重要度，建立了基于极端梯度提升算法的事故特征筛选机制，引入贝叶斯参数寻优和十折交叉验证法实现了深度森林模型的超参数优化；同时设计了循环多粒度扫描方法和加权级联森林结构，获取了交通运行风险状态等级与事故特征的映射关系，建立了基于改进深度森林模型的交通运行风险状态等级预测方法。为了验证本文方法的有效性，与支持向量机、随机森林等方法进行了对比分析，实验结果表明：本文模型预测准确度为90.80%，roc曲线下的面积 $a u c$ 值为0.99，表现出了良好的预测性能和泛化能力，与对照实验相比，本文模型具有明显的优越性，且在实效检验中同样取得了良好的预测效果。

关键词: 交通运行风险状态等级, 改进深度森林模型, 事故特征筛选, 超参数寻优

Abstract:

The prediction of traffic operation risk state level after traffic accident is the technical support for real-time risk control and emergency treatment. Based on the accident data， this paper studies the post event risk state， and establishes a traffic operation risk level prediction model based on the improved deep forest algorithm. Firstly， we analyze the importance of accident characteristics， establish the accident feature screening mechanism based on extreme gradient boosting algorithm， introduce Bayes parameter optimization and ten-fold cross validation method to realize the super parameter optimization of deep forest model. Then， we design the cyclic multi granularity scanning method and weighted cascade forest structure， obtain the mapping relationship between the traffic operation risk status level and the accident characteristics， and propose the traffic operation risk level prediction method based on the improved deep forest model. In order to verify the effectiveness of the proposed method， it is compared with SVM， RF and other methods. The experimental results show that the prediction accuracy of the model is 90.80%， and the auc value is 0.99， which show good prediction performance and generalization ability. Compared with the control experiment， the proposed model has obvious advantages. In the actual effect test， it also achieved good prediction effect.

Key words: traffic operation risk level, improved deep forest model, accident feature screening, super parameter optimization

中图分类号:

U491

李志慧,孙雅倩,陶鹏飞,李海涛,刘昕. 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), 2022, 52(1): 127-135.

Zhi-hui LI,Ya-qian SUN,Peng-fei TAO,Hai-tao LI,Xin LIU. Prediction method of traffic operation risk level after traffic accident[J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(1): 127-135.

图/表 14

图1

图2

图3

图4

图5

图6

表1

各风险等级预测结果混淆矩阵"

实际情况	预测结果
实际情况	正类	负类
正类	$T P$	$F N$
负类	$F P$	$T N$

表1

图7

图8

表2

表3

重要事故特征"

事故特征		特征变量
交通运行风险状态等级		0、1、2、3
天气状况	温度 $f 2$	-
	风寒 $f 3$	-
	湿度 $f 4$	-
	气压 $f 5$	-
	能见度 $f 6$	-
	风速 $f 8$	-
	降雨量 $f 9$	-
	天气 $f 10$	0、1、…、9
照明情况	日出日落 $f 25$	夜晚0、白天1

表3

表4

实验结果"

模型	$p e c$	$r e c$	$F 1$	$a c c$ /%
ad?gcF	0.92	0.91	0.91	90.80
ad?gcF 2	0.89	0.87	0.87	87.28
RF	0.86	0.86	0.86	86.34
SVM	0.81	0.82	0.81	81.55
ad?gcF 1	0.79	0.80	0.79	79.50

表4

图9

图10

参考文献 16

1	Kelvin K W Y, Sherrice H P L, Fung S S H. Multiple-vehicle traffic accidents in Hong Kong[J]. Accident Analysis and Prevention, 2006, 38(6): 1157-1161.
2	Çelik A K, Oktay E. A multinomial logit analysis of risk factors influencing road traffic injury severities in the Erzurum and Kars Provinces of Turkey[J]. Accident analysis and Prevention, 2014,72: 66-77.
3	Grigorios F, Achille F, Niaz G, et al. The joint effect of weather and lighting conditions on injury severities of single-vehicle accidents[J]. Analytic Methods in Accident Research,2020,27:No. 100124.
4	Fountas G, Tawfiq M S, Panagiotis C A, et al. Analysis of stationary and dynamic factors affecting highway accident occurrence: a dynamic correlated grouped random parameters binary logit approach[J]. Accident Analysis and Prevention,2018,113: 330-340.
5	孙轶轩, 邵春福, 赵丹, 等. 交通事故严重程度C5.0决策树预测模型[J]. 长安大学学报: 自然科学版, 2014, 34(5): 109-116.
	Sun Yi-xuan, Shao Chun-fu, Zhao Dan, et al. Traffic accident severity prediction model Based on C5.0 decision tree[J]. Journal of Chang'an University(Natural Science Edition), 2014, 34 (5): 109-116.
6	Burak Y K, Eren E. The novel approaches to classify cyclist accident injury-severity: hybrid fuzzy decision mechanisms.[J]. Accident Analysis and Prevention, 2020, 144: No. 105590.
7	Sameen M I, Pradhan B. Severity prediction of traffic accidents with recurrent neural networks[J]. Applied Sciences, 2017, 7(6): 432-446.
8	Amin Shohel. Backpropagation——artificial neural network (BP-ANN): Understanding gender characteristics of older driver accidents in West Midlands of United Kingdom[J]. Safety Science, 2020, 122:No. 104539.
9	Lin Y D, Li R M. Real-time traffic accidents post-impact prediction: based on crowdsourcing data[J]. Accident Analysis and Prevention, 2020, 145: 105696.
10	Bagdatli Muhammed Emin Cihangir. Vehicle delay modeling at signalized intersections with gene-expression programming[J]. Journal of Transportation Engineering, Part A: Systems,2020,146(9): No.4020107.
11	董国华, 左友兰. 交通事件下排队车辆数和总延误计算模型研究[J]. 计算技术与自动化, 2018, 37(2):110-116.
	Dong Guo-hua, Zuo You-lan. Study on calculation model of number of queuing vehicles and total delay in traffic incidents[J]. Computing Technology and Automation, 2018, 37(2): 110-116.
12	臧金蕊, 宋国华, 万涛, 等. 交通事件下快速路拥堵蔓延消散时空范围模型[J]. 交通运输系统工程与信息, 2017, 17(5): 179-185, 213.
	Zang Jin-rui, Song Guo-hua, Wan Tao, et al. A temporal and spatial model of congestion propagation and dissipation on expressway caused by traffic incidents[J]. Transportation System Engineering and Information, 2017, 17 (5): 179-185, 213.
13	商强. 基于机器学习的交通状态判别与预测方法研究[D]. 长春:吉林大学交通学院,2017.
	Shang Qiang. Research on methods for traffic state identification and prediction based on machine learning[D]. Changchun: College of Transportation, Jilin University, 2017.
14	徐铖铖, 刘攀, 王炜, 等. 恶劣天气下高速公路实时事故风险预测模型[J]. 吉林大学学报:工学版, 2013, 43(1): 68-73.
	Xu Cheng-cheng, Liu Pan, Wang Wei, et al. Real time crash risk prediction model on freeways under nasty weather conditions[J]. Journal of Jilin University(Engineering and Technology Edition), 2013, 43 (1): 68-73.
15	Fanny M, Norros I, Innamaa S. Accident risk of road and weather conditions on different road types[J]. Accident Analysis and Prevention, 2019, 122: 181-188.
16	孙轶轩, 邵春福, 岳昊, 等. 基于SVM灵敏度的城市交通事故严重程度影响因素分析[J]. 吉林大学学报: 工学版, 2014, 44(5): 1315-1320.
	Sun Yi-xuan, Shao Chun-fu, Yue Hao, et al. Urban traffic accident severity analysis based on sensitivity analysis of support vector machine[J]. Journal of Jilin University (Engineering and Technology Edition), 2014, 44 (5): 1315-1320.

相关文章 15

[1]	张文会,伊静,刘委,于秋影,王连震. 基于MADYMO的大客车追尾碰撞事故乘员损伤机理[J]. 吉林大学学报(工学版), 2022, 52(1): 118-126.
[2]	邝先验,罗会超,钟蕊,欧阳鹏. 基于天牛须小波神经网络的公交到站时间预测[J]. 吉林大学学报(工学版), 2022, 52(1): 110-117.
[3]	曲大义,黑凯先,郭海兵,贾彦峰,王韬. 车联网环境下车辆换道博弈行为及模型[J]. 吉林大学学报(工学版), 2022, 52(1): 101-109.
[4]	贾彦峰,曲大义,林璐,姚荣涵,马晓龙. 基于运行轨迹的网联混合车流速度协调控制[J]. 吉林大学学报(工学版), 2021, 51(6): 2051-2060.
[5]	杨世军,裴玉龙,潘恒彦,程国柱,张文会. 城市公交车辆驻站时间特征分析及预测[J]. 吉林大学学报(工学版), 2021, 51(6): 2031-2039.
[6]	滕文龙,丛炳虎,商云坤,张予宸,白天. 基于MEA⁃BP神经网络的建筑能耗预测模型[J]. 吉林大学学报(工学版), 2021, 51(5): 1857-1865.
[7]	陆文琦,周天,谷远利,芮一康,冉斌. 基于张量分解理论的车道级交通流数据修复算法[J]. 吉林大学学报(工学版), 2021, 51(5): 1708-1715.
[8]	李浩,陈浩. 考虑充电排队时间的电动汽车混合交通路网均衡[J]. 吉林大学学报(工学版), 2021, 51(5): 1684-1691.
[9]	马莹莹,陆思园,张晓明,魏文术. 考虑个体风险偏好差异的高速公路出行选择模型[J]. 吉林大学学报(工学版), 2021, 51(5): 1673-1683.
[10]	姚荣涵,祁文彦,郑刘杰,曲大义. 基于车道选择及行车轨迹的左转导向线设置方法[J]. 吉林大学学报(工学版), 2021, 51(5): 1651-1663.
[11]	金立生,郭柏苍,王芳荣,石健. 基于改进YOLOv3的车辆前方动态多目标检测算法[J]. 吉林大学学报(工学版), 2021, 51(4): 1427-1436.
[12]	徐进,潘存书,符经厚,刘俊,王郸祁. 典型道路场景以及场景切换时的速度行为特性[J]. 吉林大学学报(工学版), 2021, 51(4): 1331-1341.
[13]	卢凯,吴蔚,林观荣,田鑫,徐建闽. 基于KNN回归的客运枢纽聚集人数组合预测方法[J]. 吉林大学学报(工学版), 2021, 51(4): 1241-1250.
[14]	彭博,张媛媛,王玉婷,唐聚,谢济铭. 基于自动编码机-分类器的视频交通状态自动识别[J]. 吉林大学学报(工学版), 2021, 51(3): 886-892.
[15]	程国柱,程瑞,徐亮,张文会. 基于乘员伤害分析的公路路侧事故风险评价[J]. 吉林大学学报(工学版), 2021, 51(3): 875-885.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

事故特征	阈值	特征个数	准确度/%
风寒f₃	0.1662	1	46.21
温度f₂	0.1420	2	57.88
风速f₈	0.1063	3	67.42
天气f₁₀	0.1022	4	70.91
能见度f₆	0.0912	5	74.70
湿度f₄	0.0859	6	73.03
降雨量f₉	0.0741	7	79.09
气压f₅	0.0666	8	80.39
日出日落f₂₅	0.0576	9	80.81
风向f₇	0.0334	10	79.64
交通信号f₂₃	0.0127	11	79.70
人行道f₁₄	0.0104	12	78.64
交叉口f₂₆	0.0076	13	79.55