高速公路隧道出入口危险驾驶行为特性分析与识别方法

刘唐志; 潘依涵; 刘星良; 刘远强; 白致远

doi:10.3963/j.jssn.1674-4861.2025.03.005

高速公路隧道出入口危险驾驶行为特性分析与识别方法

doi: 10.3963/j.jssn.1674-4861.2025.03.005

重庆交通大学交通运输学院重庆 400074

基金项目:

国家重点研发计划项目 2023YFC3009500

国家自然科学基金项目 52302430

详细信息

作者简介:
刘唐志（1976—），博士，教授. 研究方向：交通安全、智能交通等. E-mail：liutangzhi@cqjtu.edu.cn

通讯作者:
刘星良（1989—），博士，副教授. 研究方向：交通安全、交通流理论等. E-mail: xingliang@cqjtu.edu.cn

中图分类号: U491.2
计量
- 文章访问数: 30
- HTML全文浏览量: 12
- PDF下载量: 7
- 被引次数: 0
出版历程
- 收稿日期: 2025-02-25
- 网络出版日期: 2025-10-11

An Analysis and Identification Methods of Dangerous Driving Behavior Characteristics at Highway Tunnel Entrances and Exits

College of Traffic & Transportation, Chongqing Jiaotong University, Chongqing 400074, China

摘要

摘要: 高速公路隧道出入口危险驾驶行为频发，事故风险高。针对隧道进出口区段连续轨迹数据无法有效检测导致的驾驶风险评估难题，设计了1种监测范围覆盖隧道口内外共250 m的雷达-视频融合轨迹采样系统，研究了基于特征参数优化的危险驾驶行为识别方法。基于隧道出入口轨迹数据，分析隧道出入口的驾驶行为特点，选取急变速、蛇形驾驶、高风险跟车及冒险换道4种危险驾驶行为构建危险驾驶行为谱，利用风险度量法量化4种危险驾驶行为指标，并运用四分位法设定其特征参数阈值边界，通过阈值判断，对超出阈值边界的驾驶风险点开展可视化分析，初步得出4种危险驾驶行为的空间分布特点。选用随机过采样（random oversampling，ROS）、合成少数类过采样技术（synthetic minority oversampling technique，SMOTE）、自适应综合过采样（adaptive synthetic sampling，ADASYN）对危险驾驶样本进行预处理，均衡数据样本，而后与3类集成学习算法即极端梯度提升算法（eXtreme gradient boosting，XGBoost）、轻量级梯度提升机算法（light gradient boosting machine，LGBM）、自适应提升算法（adaptive boosting，AdaBoost），通过正交组合构建均衡-集成耦合算法，共提出基于单一集成学习算法和正交组合均衡-集成算法的12种危险驾驶行为识别模型，并通过模型测试验证了不同算法模型的性能差异，确定最优危险驾驶行为识别模型。采用斯皮尔曼相关系数分析参数间的相关性，筛选出关键参数，提高模型识别性能。研究结果表明：高速公路隧道出入口因交通环境复杂性与驾驶人行为波动，成为交通事故易发路段；在3种单模态集成算法模型和9种均衡-集成耦合模型的对比评估中，基于样本优化的SMOTE-LGBM耦合模型在隧道过渡区段对危险驾驶行为的识别效果显著占优，其精确率、F-s、AUC值具体评价数值区间分别为91.2%~91.4%、0.913~0.918、0.907~0.912，相较于其他算法维持在较高水平。
- 交通安全 /
- 危险驾驶行为 /
- SMOTE-LGBM算法 /
- 隧道过渡区 /
- 驾驶行为谱
Abstract: Dangerous driving behaviors frequently occur at highway tunnel entrances and exits, posing a high risk of traffic accidents. To address the challenge of ineffective driving risk assessment caused by the inability to continuously monitor trajectory data at tunnel transition zones, this study designs a radar-video fusion trajectory sampling system with a monitoring range covering 250 meters inside and outside the tunnel portal. A dangerous driving behavior identification method based on feature parameter optimization is proposed. Based on trajectory data at tunnel entrances and exits, the characteristics of driving behavior in these zones are analyzed, and four types of dangerous driving behaviors including sudden acceleration or deceleration, serpentine driving, high-risk car-following, and aggressive lane-changing, are selected to construct a dangerous driving behavior spectrum. A risk quantification method is used to measure indicators of the four dangerous driving behaviors, and the interquartile range (IQR) method is applied to set threshold boundaries for the feature parameters. Based on these thresholds, driving risk points exceeding the boundary values are identified and visualized, and the spatial distribution characteristics of the four types of dangerous driving behaviors are preliminarily obtained. To balance the dataset, random oversampling (ROS), synthetic minority oversampling technique (SMOTE), and adaptive synthetic sampling (ADASYN) are used for sample preprocessing. Three ensemble learning methods: eXtreme gradient boosting (XGBoost), light gradient boosting machine (LGBM), and adaptive boosting (AdaBoost), are orthogonally combined with the above sampling methods to construct balanced-ensemble coupled algorithms. A total of 12 dangerous driving behavior recognition models are established, including those based on single ensemble learning algorithms and orthogonally combined balanced-ensemble algorithms. The performance differences among various models are validated through model testing to determine the optimal recognition model. Spearman correlation analysis is employed to identify key parameters and enhance model recognition performance. The research results indicate that due to the complex traffic environment and fluctuating driver behaviors, highway tunnel entrances and exits are high-risk zones for traffic accidents. Among the three single-modality ensemble models and nine balanced-ensemble coupled models evaluated, the SMOTE-LGBM coupled model based on sample optimization demonstrates superior recognition performance for dangerous driving behaviors in tunnel transition zones. Its precision, F-score, and AUC values range from 91.2% to 91.4%, 0.913 to 0.918, and 0.907 to 0.912, respectively, outperforming other algorithms and maintaining consistently high levels.
- traffic safety /
- dangerous driving behaviors /
- SMOTE-LGBM Algorithm /
- tunnel transition zones /
- driving behavior spectrum

HTML全文

图 1 隧道出入口监测路段断面划分示意图

Figure 1. Cross-sectional diagram of monitoring zones at tunnel entrance and exit

下载: 全尺寸图片幻灯片

图 2 断面区间速度变化图

Figure 2. Speed variation chart for section intervals

下载: 全尺寸图片幻灯片

图 3 断面区间加减速度分布图

Figure 3. Speed and acceleration distribution diagram of cross-sectional intervals

下载: 全尺寸图片幻灯片

图 4 换道区间位置分布图

Figure 4. Lane change interval position distribution chart

下载: 全尺寸图片幻灯片

图 5 特征参数指标危险点空间分布

Figure 5. Spatial distribution of dangerous points based on characteristic parameter indicators

下载: 全尺寸图片幻灯片

图 6 危险驾驶行为识别模型建模思路

Figure 6. Modeling approach for dangerous driving behavior recognition

下载: 全尺寸图片幻灯片

表 1 L2车道横向偏移值统计量

Table 1. Statistical metrics of L2 lane lateral offset values

统计量	隧道内横向偏移值/m	隧道外横向偏移值/m
均值	0.291 6	0.400 3
中位数	0.264 2	0.394 7
标准差	0.289 4	0.359 5
最大值	0.845 6	1.184 8
最小值	-0.305 2	-0.516 3

下载: 导出CSV

表 2 危险驾驶行为特征参数阈值

Table 2. Thresholds of characteristic parameters for dangerous driving behaviors

驾驶行为	急变速/(m/s³)	蛇形驾驶	高风险跟车/(1s)	冒险换道/(/s)
阈值	2.161\|-2.258	5.381	0.333	0.401

下载: 导出CSV

表 3 隧道出入口路段驾驶行为特征参数

Table 3. Driving behavior feature parameters for tunnel entrance and exit sections

特征参数类别	指标	含义
纵向速度参数/（m/s）	V_mean V_max V_std	速度的平均值、最大值和标准差
	△FV_mean △FV_max △FV_std	跟车速度差的平均值、最大值和标准差
	△TFV_mean △TFV_max △TFV_std	目标车道前车速度差的平均值、最大值和标准差
	△TGV_mean △TGV_max △TGV_std	目标车道后车速度差的平均值、最大值和标准差
变速参数/（m/s²）	A_mean A_max A_std	加速度的平均值、最大值和标准差
变速参数/（m/s²）	DA_mean DA_min DA_std	减速度的平均值、最小值和标准差
空间位置参数/m	D_mean D_min D_std	前车距离的平均值、最小值和标准差
	DTF_mean DTF_min DTF_std	目标车道前车距离的平均值、最小值和标准差
	DTG_mean DTG_min DTG_std	目标车道后车距离的平均值、最小值和标准差
横向动态参数/（°）	θ_mean θ_max θ_std	偏航角的平均值、最大值和标准差
横向动态参数/（°）	DLO_mean DLO_max DLO_std	横向偏移量的平均值、最大值和标准
时间占比参数/%	R_A	急加速时间占比（加速度>2.5m/s²）
	R_DA	急减速时间占比（减速度 < -2.5m/s²）
	R_V	超速时间占比（速度>80km/h）

下载: 导出CSV

表 4 混淆矩阵

Table 4. Confusion matrix

驾驶行为	识别危险行为	识别正常行为
真实危险行为	RD	TN
真实正常行为	TD	RN

下载: 导出CSV

表 5 集成学习算法评价指标

Table 5. Evaluation metrics for ensemble learning algorithms

模型类别	精确率/%			召回率/%				调和值					AUC
模型类别	D₁	D₂	D₃	D₄	D₁	D₂	D₃	D₄	D₁	D₂	D₃	D₄	D₁	D₂	D₃	D₄
AdaBoost	64.4	63.9	63.5	64.7	27.9	27.6	28.4	28.1	0.314	0.309	0.298	0.315	0.608	0.632	0.627	0.612
LGBM	38.5	37.5	38.3	39.5	23.6	23.8	22.8	23.2	0.295	0.297	0.293	0.296	0.607	0.610	0.602	0.605
XGBoost	51.1	50.9	52.8	52.7	32.2	32.1	33.1	32.7	0.411	0.407	0.413	0.409	0.664	0.671	0.667	0.668

下载: 导出CSV

表 6 危险驾驶行为识别模型性能评价指标

Table 6. Performance evaluation metrics for hazardous driving behavior recognition model

算法	精确率/%				召回率/%				调和值				AUC
算法	D₁	D₂	D₃	D₄	D₁	D₂	D₃	D₄	D₁	D₂	D₃	D₄	D₁	D₂	D₃	D₄
SMOTE-AdaBoost	84.5	83.6	84.6	84.7	87.5	87.8	87.4	88.1	0.863	0.859	0.857	0.861	0.862	0.857	0.855	0.859
SMOTE-LGBM	86.3	86.5	85.8	85.9	90.3	89.9	89.6	90.0	0.881	0.879	0.885	0.883	0.874	0.881	0.877	0.879
SMOTE-XgBoost	86.9	87.3	86.5	87.1	89.7	89.0	89.6	89.4	0.881	0.883	0.876	0.879	0.879	0.877	0.881	0.880
ADASYN-AdaBoost	82.4	82.8	82.5	81.9	89.4	89.7	89.5	89.4	0.860	0.856	0.862	0.861	0.857	0.853	0.851	0.855
ADASYN-LGBM	85.2	85.6	85.3	84.9	90.8	91.4	91.7	91.1	0.875	0.877	0.872	0.873	0.872	0.875	0.873	0.873
ADASYN-XGBoost	85.3	85.6	85.2	85.3	91.1	90.9	91.2	90.9	0.874	0.875	0.871	0.869	0.873	0.871	0.868	0.870
ROS-AdaBoost	86.9	87.3	87.2	87.5	87.4	86.8	87.5	87.2	0.884	0.886	0.877	0.883	0.886	0.883	0.885	0.883
ROS-LGBM	87.5	87.3	87.6	87.2	88.7	89.3	88.4	89.1	0.887	0.889	0.885	0.881	0.884	0.888	0.878	0.886
ROS-XGBoost	86.6	86.3	86.8	86.4	90.2	89.7	90.4	90.0	0.884	0.882	0.884	0.878	0.885	0.884	0.881	0.883

下载: 导出CSV

表 7 危险行为得分与特征参数相关系数

Table 7. Correlation coefficients between hazardous behavior scores and feature parameters

特征参数类别	相关系数
A_std	0.752
DA_std	0.736
R_A	0.341
R_DA	0.301
A_mean	0.258
DA_min	0.232
DLO_std	0.225
θ_std	0.218
DLO_mean	0.210
TG∆V_max	0.204
A_max	0.201

下载: 导出CSV

表 8 强相关性特征参数

Table 8. Highly correlated feature parameters

关联特征参数		相关系数
DLO_std	DLO_mean	0.931
DA_std	A_std	0.968
DA_std	DA_min	0.866

下载: 导出CSV

表 9 优化后的危险驾驶行为识别模型性能评价指标

Table 9. Performance evaluation metrics for the optimized hazardous driving behavior recognition model

算法	精确率/%				召回率/%					调和值				AUC
算法	D₁	D₂	D₃	D₄	D₁	D₂	D₃	D₄	D₁	D₂	D₃	D₄	D₁	D₂	D₃	D₄
AdaBoost	64.4	64.3	64.7	64.2	72.1	72.6	72.2	72.4	0.648	0.653	0.649	0.652	0.823	0.825	0.821	0.822
LGBM	53.8	54.0	53.7	53.7	68.5	67.9	67.8	68.2	0.545	0.543	0.547	0.544	0.818	0.817	0.822	0.821
XGBoost	71.1	71.4	70.8	71.3	67.3	67.1	66.9	67.1	0.738	0.737	0.741	0.740	0.842	0.838	0.843	0.841
SMOTE-AdaBoost	88.1	87.9	87.6	87.8	89.1	89.5	89.1	89.4	0.887	0.886	0.889	0.885	0.880	0.878	0.881	0.879
SMOTE-LGBM	91.2	91.4	91.3	91.2	92.3	92.2	92.5	92.6	0.917	0.915	0.918	0.913	0.912	0.909	0.907	0.910
SMOTE-XGBoost	90.9	90.5	91.0	90.8	90.5	90.1	90.4	90.4	0.911	0.913	0.909	0.911	0.906	0.907	0.906	0.905
ADASYN-AdaBoost	83.4	83.5	83.0	83.2	90.2	90.5	90.3	90.4	0.852	0.850	0.849	0.853	0.849	0.845	0.851	0.848
ADASYN-LGBM	85.2	85.0	85.5	85.3	92.8	93.2	92.9	93.0	0.870	0.869	0.873	0.871	0.868	0.865	0.868	0.867
ADASYN-XGBoost	86.1	85.8	86.2	86.0	91.3	91.7	91.5	91.6	0.890	0.891	0.887	0.892	0.879	0.882	0.877	0.881
ROS-AdaBoost	87.8	87.5	87.0	87.7	89.6	89.9	89.5	89.8	0.904	0.905	0.902	0.905	0.898	0.895	0.898	0.899
ROS-LGBM	89.8	89.5	90.0	89.6	92.6	92.9	92.5	92.8	0.907	0.905	0.908	0.904	0.905	0.903	0.903	0.906
ROS-XGBoost	88.7	88.5	88.3	88.6	91.2	91.1	90.8	91.1	0.894	0.892	0.895	0.895	0.903	0.906	0.904	0.907

下载: 导出CSV

参考文献(29)

[1]	赖金星, 张鹏, 周慧, 等. 高速公路隧道交通事故规律研究[J]. 隧道建设, 2017, 37(1): 37-42. LAI J X, ZHANG P, ZHOU H, et al. Study on traffic accident patterns in highway tunnels[J]. Tunnel Construction, 2017, 37 (1), 37-42. (in Chinese)
[2]	李晓赫. 基于无人机航拍视频的车辆运动数据提取方法研究[D]. 北京: 清华大学, 2023. LI X H. Research on vehicle motion data extraction methods based on drone aerial video[D]. Beijing: Tsinghua University, 2023. (in Chinese)
[3]	黄海南, 沈正航, 徐锦强, 等. 基于无人机视频采集的车辆驾驶行为分析实验设计[J]. 实验技术与管理, 2023, 40(10): 85-90. HUANG H N, SHEN Z H, XU J Q, et al. Experimental design for vehicle driving behavior analysis based on drone video capture[J]. Experimental Technology and Management, 2023, 40(10), 85-90. (in Chinese)
[4]	薛清文, 蒋愚明, 陆键. 基于轨迹数据的危险驾驶行为识别方法[J]. 中国公路学报, 2020, 33(6): 84-94. XUE Q. W, JIANG Y M, LU J. Dangerous driving behavior recognition method based on trajectory data[J]. China Journal of Highway and Transport, 2020, 33(6): 84-94. (in Chinese)
[5]	王雪松, 朱美新. 基于自然驾驶数据的中国驾驶人城市快速路跟驰模型标定与验证[J]. 中国公路学报, 2018, 31(9): 129-137. WANG X S, ZHU M X. Calibration and validation of urban expressway car-following model based on natural driving data in China[J]. China Journal of Highway and Transport, 2018, 31(9), 129-137. (in Chinese)
[6]	王可, 陆键, 蒋愚明. 基于车辆行驶轨迹的道路不良驾驶行为谱构建与特征值计算方法[J]. 交通运输工程学报, 2020, 20 (6): 236-249. WANG K, LU J, JIANG Y M. Construction of poor driving behavior spectrum and feature value calculation method based on vehicle trajectory[J]. Journal of Traffic and Transportation Engineering, 2020, 20(6), 236-249. (in Chinese)
[7]	刘唐志, 毕辉云, 杨卓思, 等. 基于操纵量指标的合流区危险驾驶行为谱研究[J]. 交通运输系统工程与信息, 2023, 23 (2): 242-251. LIU T Z, BI H Y, YANG Z S, et al. Study on dangerous driving behavior spectrum in merging areas based on maneuvering indicators[J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(2): 242-251. (in Chinese)
[8]	陈曦, 赵欣, 林友欣, 等. 基于高精度轨迹的交叉口急变速行为识别研究[J]. 武汉理工大学学报(交通科学与工程版), 2023, 47(1): 61-66, 72. CHEN X, ZHAO X, LIN Y X, et al. Study on sudden speed change behavior recognition at intersections based on high-precision trajectories[J]. Journal of Wuhan University of Technology(Traffic Science and Engineering), 2023, 47 (1), 61-66, 72. (in Chinese)
[9]	张雅楠, 唐阳山, 刘昊, 等. 基于行车数据的驾驶人危险驾驶行为的判定[J]. 汽车实用技术, 2019, (15): 247-248. ZHANG Y N, TANG Y S, LIU H, et al. Determination of dangerous driving behavior of drivers based on driving data[J]. Automobile Applied Technology, 2019, (15), 247-248. (in Chinese)
[10]	公安部交通管理局. 中华人民共和国道路交通事故统计年报(2016-2019年度)[Z]. 北京: 公安部交通管理局, 2020. Ministry of Public Security Traffic Management Bureau. Road traffic accident statistical yearbook of the people's republic of China(2016-2019)[Z]. Beijing: Traffic Management Bureau, Ministry of Public Security, 2020. (in Chinese)
[11]	龙彦, 黄建玲, 赵晓华. 换道过程中驾驶人感知操作的模式发现与规则挖掘[J]. 交通运输系统工程与信息, 2021, 21 (3): 237-246. LONG Y, HUANG J L, ZHAO X H. Pattern discovery and rule mining of drivers' perception operations during lane-changing process[J]. Journal of Transportation Systems Engineering and Information Technology, 2021, 21(3): 237-246. (in Chinese)
[12]	贺玉龙, 刘磊, 迟佳欣. 高速公路车辆换道行为风险研究[J]. 重庆交通大学学报(自然科学版), 2021, 40(4): 26-33. HE Y L, LIU L, CHI J X. Study on the risk of lane change behavior on highways[J]. Journal of Chongqing Jiaotong University(Natural Science), 2021, 40(4), 26-33. (in Chinese)
[13]	SHI X, WONG Y D, LI M, et al. A feature learning approach based on XGBoost for driving assessment and risk prediction[J]. Accident Analysis & Prevention, 2019, 129: 17-179.
[14]	车翔玖, 于英杰, 刘全乐. 增强Bagging集成学习及多目标检测算法[J]. 吉林大学学报(工学版), 2022, 52(12): 2916-2923. CHE X J, YU Y J, LIU Q L. Enhanced bagging ensemble earning and multi-objective detection algorithm[J]. Journal of Jilin University (Engineering and Technology Edition), 2022, 52(12), 2916-2923. (in Chinese)
[15]	杨晓菡, 郝国生, 张谢华, 等. 基于协同训练与Boosting的协同过滤算法[J]. 计算机应用, 2023, 43(10): 3136-3141. YANG X H, HAO G S, ZHANG X H, et al. Collaborative filtering algorithm based on co-training and Boosting[J]. Journal of Computer Applications, 2023, 43(10), 3136-3141. (in Chinese)
[16]	TANG J, LIU G, PAN Q. A review on representative swarm intelligence algorithms for solving optimization problems: applications and trends[J]. IEEE/CAA Journal of Automatica Sinica, 2021(10): 1627-1643.
[17]	WANG W, SUN D. The improved AdaBoost algorithms for imbalanced data classification[J]. Information Sciences, 2021, 563: 358-374. doi: 10.1016/j.ins.2021.03.042
[18]	王建军, 李冬怡, 王赛, 等. 基于EWM-TOPSIS的城市卡口地点车速数据诊断[J]. 长安大学学报(自然科学版), 2023, 43 (3): 67-75. WANG J J, LI D Y, WANG S, et al. Vehicle speed data diagnosis of urban checkpoint locations based on EWM-TOPSIS[J]. Journal of Chang'an University(Natural Science Edition), 2023, 43(3), 67-75. (in Chinese)
[19]	张雷元, 张韧, 刘海平. 基于卡口监测数据的路段交通异常状态识别方法研究[J]. 中国人民公安大学学报(自然科学版), 2023, 29(1): 83-87. ZHANG L Y, ZHANG R, LIU H P. Research on the method for identifying traffic anomalies on road sections based on checkpoint monitoring data[J]. Journal of People's Public Security University of China (Science and Technology), 2023, 29(1), 83-87. (in Chinese)
[20]	林译峰, 温惠英. 高速公路隧道不同车辆实时跟驰风险影响因素分析[J]. 交通运输系统工程与信息, 2025, 25(1): 298-310. LIN Y F, WEN H Y. Analysis of factors influencing real time following risk of different vehicles in highway tunnels[J]. Journal of Transportation Systems Engineering and Information Technology, 2025, 25(1): 298-310. (in Chinese)
[21]	赵天辉, 张耀, 王建学. 基于空间密度聚类和异常数据域的负荷异常值识别方法[J]. 电力系统自动化, 2021, 45(10): 97-105. ZHAO T H, ZHANG Y, WANG J X. Load outlier identification method based on spatial density clustering and anomalous data domain[J]. Automation of Electric Power Systems, 2021, 45(10): 97-105. (in Chinese)
[22]	王伟, 赵琦, 王力, 等. 基于车辆轨迹数据的急减速驾驶行为判定方法[J]. 科学技术与工程, 2022, 22(10): 4215-4221. WANG W, ZHAO Q, WANG L, et al. Method for determining sudden deceleration driving behavior based on vehicle trajectory data[J]. Science Technology and Engineering, 2022, 22(10), 4215-4221. (in Chinese)
[23]	金辉, 吕明. 基于得分系数的跟车工况驾驶风格识别研究[J]. 北京理工大学学报, 2021, 41(3): 245-250. JIN H, LYU M. Research on car-following driving style recognition based on scoring coefficient[J]. Transactions of Bejing Institute of Technology, 2021, 41(3), 245-250. (in Chinese)
[24]	吴斌, 朱西产, 沈剑平. 基于自然驾驶数据的驾驶人紧急转向变道模型[J]. 同济大学学报(自然科学版), 2019, 47 (11): 1618-1625. WU B, ZHU X C, SHEN J P. Emergency steering lane change model based on natural driving data[J]. Journal of Tongji University(Natural Science), 2019, 47(11), 1618-1625. (in Chinese)
[25]	CHEN S, XUE Q, ZHAO X, et al. Risky driving behavior recognition based on vehicle trajectory[J]. International Journal of Environmental Research and Public Health, 2021, 18 (23): 12373-12373. doi: 10.3390/ijerph182312373
[26]	CHEN S, CHENG K, YANG J, et al. Driving behavior risk measurement and cluster analysis driven by vehicle trajectory data[J]. Applied Sciences, 2023, 13(9): 5675. doi: 10.3390/app13095675
[27]	DAS S, MAURYA A K. Defining time-to-collision thresholds by the type of lead vehicle in nor-lane-based traffic environment[J]. IEEE Transactions on Intelligent Transportation Systems, 2019, 21(12): 1-11.
[28]	谢厅, 刘星良, 刘唐志, 等. 基于DE-EL的城市快速路合流区危险驾驶行为识别方法[J]. 交通信息与安全, 2024, 42 (6): 23-30. doi: 10.3963/j.jssn.1674-4861.2024.06.003 XIE T, LIU X L, LIU T Z, et al. A method for identifying dangerous driving behaviors in urban expressway merging areas based on DE-EL[J]. Journal of Transport Information and Safety, 2024, 42(6): 23-30. (in Chinese) doi: 10.3963/j.jssn.1674-4861.2024.06.003
[29]	杨亚柳. 基于ARIMA模型和ARDL模型对COVID-19疫情的研究[D]. 哈尔滨: 哈尔滨工业大学, 2021. YANG Y L. Research on COVID-19 epidemic based on ARIMA model and ARDL model[D]. Harbin: Harbin Institute of Technology, 2021. (in Chinese)