基于多分类监督学习的驾驶风格特征指标筛选

王旭; 马菲; 廖小棱; 蒋佩玉; 张伟; 王芳

doi:10.3963/j.jssn.1674-4861.2022.01.019

基于多分类监督学习的驾驶风格特征指标筛选

doi: 10.3963/j.jssn.1674-4861.2022.01.019

王旭^1, ,,
马菲¹,
廖小棱¹,
蒋佩玉¹,
张伟^{2, 3},
王芳⁴

1.
山东大学齐鲁交通学院济南 250001
2.
山东高速集团有限公司济南 250001
3.
山东省智慧交通重点实验室济南 250001
4.
山东高速信息集团有限公司济南 250001

基金项目:

山东省重点研发计划项目 2020CXGC010117

山东省交通科技计划项目 2021B60

详细信息

通讯作者:
王旭（1987—），博士，副教授.研究方向：智能交通系统、主动交通控制管理、道路交通安全.Email：xuwang@sdu.edu.cn

中图分类号: U491.6
计量
- 文章访问数: 1167
- HTML全文浏览量: 445
- PDF下载量: 79
- 被引次数: 0
出版历程
- 收稿日期: 2021-09-23
- 网络出版日期: 2022-03-31

Feature Selection for Recognition of Driving Styles Based on Multi-Classification and Supervised Learning

WANG Xu^{1
, ,},
MA Fei¹,
LIAO Xiaoling¹,
JIANG Peiyu¹,
ZHANG Wei^{2, 3},
WANG Fang⁴

1.
School of Qilu Transportation, Shandong University, Jinan 250001, China
2.
Shandong Hi-Speed Group Co., Ltd., Jinan 250001, China
3.
Shandong Key Laboratory of Intelligent Transportation, Jinan 250001, China
4.
Shandong Hi-Speed Information Group Co., Ltd., Jinan 250001, China

摘要

摘要:
交通事故与驾驶风格具有强烈的相关性，而驾驶风格的直观体现是驾驶行为。为深入分析驾驶行为与驾驶风格的关联性，探索不同驾驶风格群体之间的差异，筛选驾驶风格分类与识别影响因素，建立驾驶风格识别模型并验证有效性。依托车联网实验数据，利用K-means++算法对驾驶员样本数据集进行驾驶风格聚类，设计支持向量机-递归特征消除（SVC-RFE）与随机森林-递归特征消除（RF-RFE）算法进行驾驶特征重要度排序，利用筛选出的特征指标搭建神经网络驾驶风格识别模型。结果表明：①特征个数n = 6时，2种特征排序算法的排序正确率均高于85%，其中RF-RFE的正确率可达90%；②特征排序中重要度最高的指标为最大速度，其在3种驾驶风格群体中的差异可达10 m/s；③仅以最大速度作为输入，驾驶风格识别模型精度为86.1%，表明最大速度可有效区分驾驶风格。
- 交通工程 /
- 特征排序 /
- 递归特征消元 /
- 驾驶风格 /
- 交通安全
Abstract:
Traffic accidents are strongly correlated with driving style, and driving style can be intuitively represented by driving behavior. In order to further advance understanding of the relationship between driving behavior and driving style, this paper explores thedifferences between driving styles and identifies factors that affect the classification. A driving-style recognition model is then proposed and evaluated. Based on the experimental data from connected vehicles, a K-means++ algorithm is proposed and used to classify data of driving behavior under different driving styles and a support vector machine-recursive feature elimination(SVC-RFE)and a random forest-recursive feature elimination(RF-RFE)algorithm are used to rank the importance of features of driving behavior. A classification model for driving styles based on neural network and the above selected features is developed. The results show that: ①when the number of selected features is set as n = 6, the correct ranking rate of both feature ranking algorithms is above 85% and the correct rate of the RF-RFEalgorithm is up to 90%.②The indicator with the highest importance in feature ranking is the maximum speed, and its difference among the three driving style groups is up to 10 m/s. ③When only the maximum speed is used as input, the accuracy of the driving-style recognition model is 86.1% and therefore, it can be concluded that maximum speed can effectively distinguish driving styles.
- traffic engineering /
- feature sorting /
- recursive feature elimination /
- driving style /
- traffic safety

HTML全文

图 1 主成分贡献率图

Figure 1. Principal Component Contribution Rate

下载: 全尺寸图片幻灯片

图 2 特征个数与交叉验证正确分类分值

Figure 2. The Number of Features and The Correct Classification Score of Cross Validation

下载: 全尺寸图片幻灯片

图 3 最大速度区间

Figure 3. Maximum speed interval diagram

下载: 全尺寸图片幻灯片

图 4 平均速度区间

Figure 4. Average speed interval diagram

下载: 全尺寸图片幻灯片

图 5 测试样本测试结果

Figure 5. Test sample test results

下载: 全尺寸图片幻灯片

表 1 驾驶风格量化指标集

Table 1. Driving style quantitative index set

序号	评价指标	序号	评价指标
1	平均车速/（m/s）	10	最小横向加速度/（m/s²）
2	最大车速/（m/s）	11	平均垂向加速度/（m/s²）
3	最小车速/（m/s）	12	最大垂向加速度/（m/s²）
4	速度标准差/（m/s）	13	最小垂向加速度/（m/s²）
5	平均纵向加速度/（m/s²）	14	平均偏航率/（（°）/s）
6	最大纵向加速度/（m/s²）	15	偏航率标准差/（（°）/s）
7	最小纵向加速度/（m/s²）	16	纵向冲击度平均值/（m/s³）
8	平均横向加速度/（m/s²）	17	垂向冲击度平均值/（m/s³）
9	最大横向加速度/（m/s²）	18	行程距离/m

下载: 导出CSV

表 2 各主成分的信息贡献率与累积贡献率

Table 2. Information Contribution Rate and Cumulative Contribution Rate of Each Principal Component 单位: %

主成分	信息贡献率b_j	累积贡献率α
第1主成分	24.2	24.2
第2主成分	21.9	46.1
第3主成分	15.0	61.1
第4主成分	12.5	73.6
第5主成分	7.4	81.0
第6主成分	4.0	85.0

下载: 导出CSV

表 3 驾驶员主成分得分

Table 3. Driver principal component score

测试样本	第1 主成分	第2 主成分	第3 主成分	第4 主成分	第5 主成分	第6 主成分
样本1	-0.147	-0.738	-1.055	-1.664	-10.172	6.380
样本2	0.097	-0.814	-0.964	-1.235	-2.273	2.066
︙	︙	︙	︙	︙	︙	︙
样本241	-2.667	0.165	4.557	-0.116	-9.724	2.030
样本242	0.325	2.938	3.062	2.356	-13.067	8.413

下载: 导出CSV

表 4 各指标的因子载荷量

Table 4. Factor load of each index

指标	第1 主成分	第2 主成分	第3 主成分	第4 主成分	第5 主成分	第6 主成分
指标1	-0.399	-0.005	0.007	0.875	-0.015	0.001
指标2	-0.559	0.420	-0.133	0.608	0.237	-0.034
指标3	0.309	-0.565	0.275	0.508	-0.291	-0.169
指标4	-0.428	0.676	-0.220	-0.014	0.436	0.006
指标5	0.246	-0.284	-0.513	0.240	0.023	-0.137
指标6	-0.389	0.510	-0.412	-0.059	-0.342	-0.253
指标7	0.394	-0.525	0.256	0.170	0.404	0.254
指标8	0.705	0.639	-0.005	0.210	-0.175	-0.022
指标9	0.696	0.646	-0.010	0.218	-0.172	0.003
指标10	0.710	0.632	0.000	0.208	-0.173	-0.049
指标11	0.669	0.404	0.114	0.225	0.356	-0.100
指标12	-0.178	-0.357	0.112	0.047	0.151	-0.836
指标13	0.586	-0.228	0.279	0.127	0.493	-0.083
指标14	-0.371	0.384	0.653	0.007	0.033	-0.084
指标15	0.131	0.515	-0.212	-0.497	0.369	-0.184
指标16	-0.335	0.366	0.841	-0.075	-0.048	-0.018
指标17	0.228	-0.262	-0.833	0.040	0.106	-0.008
指标18	-0.647	0.185	-0.226	0.392	0.165	0.194

下载: 导出CSV

表 5 K-means与K-means++聚类效果

Table 5. Clustering effect of K-means and K-means++

方法类型	聚类数K	迭代次数	轮廓系数
K-means	3	16	0.339 4762
K-means++	3	16	0.349 6989

下载: 导出CSV

表 6 SVC-RFE与RF-RFE排名前6的特征及重要性分值

Table 6. Top 6 characteristics and importance scores of SVC-RFE and RF-RFE

SVC-RFE		RF-RFE
特征变量	重要度分值	特征变量	重要度分值
平均纵向加速度	4.979	最大速度	0.086 7
垂向加速度平均值	2.290	最小垂向加速度	0.021 0
速度标准差	1.786	最小纵向加速度	0.020 5
最小纵向加速度	0.909	速度标准差	0.014 6
最小垂向加速度	0.706	距离	0.012 9
最大速度	0.207	平均速度	0.009 0

下载: 导出CSV

参考文献(24)

[1]	国家统计局. 中国统计年鉴2019[M]: 中国统计出版社, 2020. National Bureau of Statistics of China. China statistical yearbook 2019[M]: Beijing: China Statistics Press, 2020. (in Chinese)
[2]	郭孜政. 驾驶行为险态辨识理论与方法[D]. 成都: 西南交通大学, 2009. GUO Z Z. Theories and methods on driving risk status identification[D]. Chengdu: Southwest Jiaotong University, 2009. (in Chinese)
[3]	American Automobile Association. Aggressive driving: Research update[J]. American Automobile Association Foundation for Traffic Safety, 2009: 5-6.
[4]	CARSTEN O, KIRCHER K, JAMSON S. Vehicle-based studies of driving in the real world: The hard truth?[J]. Accident Analysis & Prevention, 2013(58): 162-74.
[5]	侯海晶, 金立生, 关志伟, 等. 驾驶风格对驾驶行为的影响[J]. 中国公路学报, 2018, 31(4): 18-27. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL201804004.htm HOU H J, JIN L S, GUAN Z W, et al. Effects of driving style on driver behavior[J]. China Journal of Highway and Transport, 2018, 31(4): 18-27. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL201804004.htm
[6]	YI D W, SU J Y, LIU C J, et al. A machine learning based personalized system for driving state recognition[J]. Transportation Research Part C: Emerging Technologies, 2019, 105: 241-261. doi: 10.1016/j.trc.2019.05.042
[7]	OUALI T, SHAH N, KIM B, et al. Driving style identification algorithm with real-world data based on satistical approach[C]. SAE 2016 World Congress and Exhibition, Detroit, Michigan, U.S. : SAE International, 2016.
[8]	孙龙, 常若松. 驾驶风格研究现状与展望[J]. 人类工效学, 2013, 19(4): 92-95. https://www.cnki.com.cn/Article/CJFDTOTAL-XIAO201304024.htm SUN L, CHANG R S. Research status and prospect of driving style[J]. Ergonomics, 2013, 19(4): 92-95. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-XIAO201304024.htm
[9]	WANG X, XU X. Assessing the relationship between self-reported driving behaviors and driver risk using a naturalistic driving study[J]. Accident Analysis & Prevention, 2019 (128): 8-16.
[10]	PADILLA J L, CASTRO C, DONCEL P, et al. Adaptation of the multidimensional driving styles inventory for Spanish drivers: convergent and predictive validity evidence for detecting safe and unsafe driving styles[J]. Accident Analysis & Prevention, 2020, 136(C): 105413.
[11]	BELLEM H, SCHöNENBERG T, KREMS J F, et al. Objective metrics of comfort: Developing a driving style for highly automated vehicles[J]. Transportation Research Part F: Traffic Psychology and Behaviour, 2016(41): 45-54.
[12]	CASTIGNANI G, DERRMANN T, FRANK R, et al. Driver behavior profiling using smartphones: A low-cost platform for driver monitoring[J]. IEEE Intelligent Transportation Systems Magazine, 2015, 7(1): 91-102. doi: 10.1109/MITS.2014.2328673
[13]	KARGINOVA N, BYTTNER S, SVENSSON M. Data-driven methods for classification of driving styles in buses[C]. SAE 2012 World Congress and Exhibition, Detroit, Michigan, U.S. : SAE International, 2012.
[14]	ASTARITA V, FESTA D C, GIOFRÈ, P. et al. Co-operative ITS: ESD a smartphone based system for sustainability and transportation safety[J]. Procedia Computer Science, 2016 (83): 449-456.
[15]	吕能超, 任泽远, 段至诚, 等. Near-crash事件中驾驶人行为特征分析[J]. 中国安全科学学报, 2017, 27(6): 19-24. https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK201706004.htm LYU N C, REN Z Y, DUAN Z C, et al. Influencing factors of driving risk based on critical incident events[J]. Journal of Transport Information and Safety, 2017, 27(6): 19-24. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZAQK201706004.htm
[16]	MA Y F, LI W L, TANG K, et al. Driving style recognition and comparisons among driving tasks based on driver behavior in the online car-hailing industry[J]. Accident Analysis & Prevention, 2021(154): 1-12.
[17]	杨曼, 吴超仲, 张晖, 等. 行车安全事件的驾驶风险影响因素研究[J]. 交通信息与安全, 2018, 36(5): 34-39. doi: 10.3963/j.issn.1674-4861.2018.05.005 YANG M, WU C Z, ZHANG H, et al. Influencing factors of driving risk based on critical incident events[J]. Journal of Transport Information and Safety, 2018, 36(5): 34-39. (in Chinese) doi: 10.3963/j.issn.1674-4861.2018.05.005
[18]	GIANNA C, HEIMBRAND S, GRESTY M. Thresholds for detection of motion direction during passive lateral whole-body acceleration in normal subjects and patients with bilateral loss of labyrinthine function[J]. Brain Research Bulletin, 1996, 40(5/6): 443.
[19]	李立治, 杨建军, 刘双喜, 等. 国内人群的驾驶风格分类及识别方法研究[J]. 重庆理工大学学报(自然科学), 2019, 33 (11): 33-40. https://www.cnki.com.cn/Article/CJFDTOTAL-CGGL201911005.htm LI L Z, YANG J J, LIU S X, et al. Research on classification and recognition of driving style of domestic crowds[J]. Journal of Chongqing University of Technology(Natural Science), 2019, 33(11): 33-40. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-CGGL201911005.htm
[20]	杨俊闯, 赵超. k-means聚类算法研究综述[J]. 计算机工程与应用, 2019, 55(23): 7-14+63. https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG201923003.htm YANG J C, ZHAO C. Survey on K-means clustering algorithm[J]. Computer Engineering and Applications, 2019, 55 (23): 7-14+63. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG201923003.htm
[21]	YU G, WENG G R. K-means++ clustering-based active contour model for fast image segmentation[J]. Journal of Electronic Imaging, 2018, 27(6): 1-12.
[22]	GYS A M M, HERMANUS C M. A review of intelligent driving style analysis systems and related artificial intelligence algorithms[J]. Sensors(Basel, Switzerland), 2015, 15 (12): 30653-30682.
[23]	吴辰文, 梁靖涵, 王伟, 等. 基于递归特征消除方法的随机森林算法[J]. 统计与决策, 2017(21): 60-63. https://www.cnki.com.cn/Article/CJFDTOTAL-TJJC201721016.htm WU C W, LIANG J H, WANG W, et al. Random forest algorithm based on recursive feature elimination[J]. Statistics and Decision Making, 2017(21): 60-63. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-TJJC201721016.htm
[24]	叶明全, 高凌云, 伍长荣, 等. 基于对称不确定性和SVM递归特征消除的信息基因选择方法[J]. 模式识别与人工智能, 2017, 30(5): 429-438. https://www.cnki.com.cn/Article/CJFDTOTAL-MSSB201705005.htm YE M Q, GAO L Y, WU C R, et al. Informative gene selection method based on symmetric uncertainty and SVM recursive feature elimination[J]. Pattern Recognition and Artificial Intelligence, 2017, 30(5): 429-438. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-MSSB201705005.htm