An Analysis of Traffic Accident Severity Based on Mutual-information Bayesian Network
-
摘要: 为掌握省际客运行业事故严重程度影响因素, 采用互信息及贝叶斯网络方法构建模型, 分析各因素变化与事故严重程度的定量互动关系。鉴于行业样本量较小及专家知识建模存在主观性, 采用改进离散算法挖掘数据, 提出结合互信息与交叉验证的先验网络构造方法。以上海市2005—2019年741起省际客运事故数据为例进行模型分析。结果表明: 对事故最敏感的影响因素为驾驶员性别、天气和车辆类型; 其中“女性驾驶员”“雪、大风、雾”“中型客车”对事故严重性的权重占比分别为13.5%, 8.8%和5.7%;此外, 驾驶员年龄对群死群伤事故贡献较小; 客车尺寸与安全性非单调关系; 00:00—05:00引发7人以上受伤的概率同比上升9%;季节、天气、时间因素与财产损失无直接关联。模型泛化能力优于对比模型, AUC均值为0.644 588, 命中率达到97.3%。Abstract: The methods of mutual information and Bayesian network are conducted to develop a model to grasp the factors affecting the severity of accidents in the inter-provincial bus industry. The quantitative interaction between changes in factors and the severity of accidents are analyzed. Given the limitation of the samples' size of the industry and the subjectivity of experts' knowledge of modeling, an improved discrete algorithm is used for data mining.A primary network construction method combining mutual information and cross-validation is proposed. Taking model analysis with 741 inter-provincial bus accidents in Shanghai from 2005 to 2019 as a case study, the results show that the most sensitive influencing factors of accidents are gender, weather, and vehicle type."Female driver""snow, wind, and fog""medium-size bus"account for 13.5%, 8.8%, and 5.7% of the weight of the accidents, respectively. Additionally, drivers' age has little contribution to the misfortune of group death and injury. Bus size has non-monotonic relationships with safety. The probability of more than seven people being injured during 00:00 to05:00 rises by 9%. The factors of season, weather, and time are not directly related to property loss. The generalization ability of the constructed model is better than other comparable models. The average AUC is 0.644 588, and the hit rate reaches 97.3%.
-
Key words:
- traffic safety /
- inter-provincial bus /
- accident analysis /
- Bayesian network /
- mutual information
-
图 5 省际客运事故严重程度分析模型
Figure 5. Accident severity analysis model of inter-provincial passenger transport
表 1 建模变量区间划分
Table 1. Variable-interval division of modeling
变量名称 区间（类）划分 变量名称 区间（类）划分 性别 男—① 行驶状态* 直行—① 女—② 停车、倒车、掉头、左转弯—② 年龄* 0~27—① 右转弯—③ 28~47—② 变更车道、躲避障碍、驶离路面—④ 48~51—③ 其他—⑤ 52~54—④ 死亡人数* 0 —① ＞54 —⑤ 1 —② 号牌种类 大型汽车—① 2 —③ 小型汽车—② 3 —④ 车辆类型 大型客车—① ＞3 —⑤ 中型客车—② 重伤人数* 无—① 小型客车—③ 有—② 事故地点 路口—① 轻伤人数* 0 —① 路段—② 0~7 —② 季节 春—① 8~13—③ 夏—② 14~25—④ 秋—③ ＞25 —⑤ 冬—④ 财产损失* 0~110 —① 天气* 晴—① 111~515.5 —② 阴—② 516.5~9 185 —③ 雨—③ ＞9 185 —④ 雪、大风、雾—④ 事故类型 死亡事故—① 时间 00：00~05：00—① 受伤事故—② 06：00~11：00—② 财产损失事故—③ 12：00~17：00—③ 18：00~23：00—④ 注：*号表示经过CACC离散处理的变量。 
下载: 导出CSV
表 2 互信息矩阵
Table 2. Mutual information matrix
因素 事故地点 重伤人数 死亡人数 号牌种类 性别 季节 轻伤人数 时间 车辆类型 天气 财产损失 行驶状态 年龄 事故类型 事故地点 10.78 10.65 10.78 10.63 10.78 10.71 10.78 10.38 10.64 10.78 10.07 9.93 10.78 重伤人数 10.78 10.58 10.78 10.63 10.78 10.71 10.78 10.38 10.63 10.36 10.22 10.38 10.66 死亡人数 10.65 10.58 10.53 10.49 10.11 10.37 10.48 10.47 10.07 9.96 10.31 10.12 10.64 号牌种类 10.78 10.78 10.53 10.63 10.78 10.71 10.78 10.78 10.64 10.78 10.25 10.51 10.64 性别 10.63 10.63 10.49 10.63 10.64 10.55 10.63 10.62 10.48 10.63 10.55 10.62 10.62 季节 10.78 10.78 10.11 10.78 10.64 10.71 9.99 10.41 9.75 10.08 8.78 8.80 10.51 轻伤人数 10.71 10.71 10.37 10.71 10.55 10.71 10.71 10.31 10.58 10.71 10.18 9.73 10.39 时间 10.78 10.78 10.48 10.78 10.63 9.99 10.71 10.40 10.21 9.85 9.05 8.95 10.30 车辆类型 10.38 10.38 10.47 10.78 10.62 10.41 10.31 10.40 10.26 10.40 9.60 10.16 10.25 天气 10.64 10.63 10.07 10.64 10.48 9.75 10.58 10.21 10.26 9.69 9.52 9.40 9.97 财产损失 10.78 10.36 9.96 10.78 10.63 10.08 10.71 9.85 10.40 9.69 9.31 9.60 9.94 行驶状态 10.07 10.22 10.31 10.25 10.55 8.78 10.18 9.05 9.60 9.52 9.31 10.01 9.66 年龄 9.93 10.38 10.12 10.51 10.62 8.80 9.73 8.95 10.16 9.40 9.60 10.01 9.64 事故类型 10.78 10.66 10.64 10.64 10.62 10.51 10.39 10.30 10.25 9.97 9.94 9.66 9.64
下载: 导出CSV
表 3 变量节点序列
Table 3. Variable node sequence
序号 变量 MI值 描述 0 事故类型 结果变量 1 重伤人数 10.661 8 2 死亡人数 10.645 9 3 轻伤人数 10.396 8 4 财产损失 9.941 4 5 事故地点 10.780 9 影响因素变量 6 号牌种类 10.640 8 7 性别 10.625 5 8 季节 10.512 7 9 时间 10.306 9 10 车辆类型 10.250 3 11 天气 9.971 8 12 行驶状态 9.667 4 13 年龄 9.647 8
下载: 导出CSV
表 4 条件概率分析
Table 4. Conditional probability analysis
因素名称 类 比例/% 死亡事故/% 死亡人数-⑤/% 平均值/% 归一化/% 敏感度归一化/% 事故影响权重/% 基准值 18.0 5.0 性别 ① 98.2 18.0 4.0 11.0 31.0 18.0 6.1 ② 1.8 32.0 17.0 24.5 69.0 32.0 13.5 车辆类型 ① 53.9 22.0 5.0 13.5 36.0 22.0 4.8 ② 3.9 26.0 6.0 16.0 42.7 26.0 5.7 ③ 42.2 12.0 4.0 8.0 21.3 12.0 2.8 天气 ① 66.9 17.0 3.0 10.0 17.2 17.0 4.2 ② 15.3 19.0 8.0 13.5 23.3 19.0 5.7 ③ 16.4 20.0 7.0 13.5 23.3 20.0 5.7 ④ 1.5 26.0 16.0 21.0 36.2 26.0 8.8
下载: 导出CSV
表 5 死亡人数后验概率分析
Table 5. Posterior probability analysis of the number of deaths
影响因素 类 死亡人数概率变化/% 0 1 2 3 ＞3 性别 ② 29↓ 2↑ 8↑ 8↑ 12↑ ③ 1↓ 1↑ - 年龄 ④ 3↓ 4↑ 1↓ ⑤ 4↓ 5↑ 1↓ 号牌种类 ① 5↓ 5↑ 1↓ 车辆类型 ① 5↓ 6↑ 1↓ ② 16↓ 9↓ 3↑ 2↑ 1↑ 事故地点 ② 5↓ 4↑ 1↑ 季节 ③ 2↓ 1↑ ④ 2↓ 1↑ 1↑ 天气 ② 6↓ 1↓ 2↑ 1↑ 3↑ ③ 7↓ 1↑ 2↑ 2↑ 2↑ ④ 24↓ 1↑ 6↑ 6↑ 11↑ 时间 ① 5↓ - 1↑ 1↑ 3↑ 行驶状态 ② 3↓ 1↑ 1↑ 1↑ 1↑ ⑤ 11↓ 2↑ 3↑ 3↑ 3↑
下载: 导出CSV
表 6 受伤及财产损失后验概率分析
Table 6. Posterior probability analysis of injury and property loss
影响因素 类 轻伤人数概率变化/% 重伤人数概率变化/% 财产损失概率变化/% 0~7 8~13 14~25 ＞25 有 0~110 111~515.5 516.5~9185 ＞9185 性别 ② 14↑ 14↑ 14↑ 14↑ 19↑ 2↑ 11↓ 9↑ ① 4↑ 1↑ 4↑ 6↑ 1↑ 6↓ 1↓ 年龄 ③ 1↓ 2↓ 2↑ ④ 2↓ 3↓ 5↓ 1↓ 5↑ 1↑ ⑤ 3↓ 4↓ 7↓ 1↓ 7↑ 1↑ 号牌种类 ① 4↓ 5↓ 8↓ 2↓ 8↑ 1↑ ② 6↑ 1↑ 1↑ 1↑ 6↑ 10↑ 2↑ 10↓ 2↓ ① 4↓ 5↓ 8↓ 1↓ 8↑ 1↑ 车辆类型 ② 3↓ 4↓ 7↓ 1↓ 7↑ 1↑ ③ 6↑ 1↑ 1↑ 1↑ 6↑ 10↑ 2↑ 10↓ 2↓ 事故地点 ① 2↑ 3↑ 4↑ 2↑ 2↓ 4↓ ② 2↓ 4↓ 5↓ 2↓ 3↑ 4↑ 季节 ③ 1↑ 1↑ 2↑ 1↑ 1↑ ④ 1↓ 1↑ 1↑ 1↑ 3↑ ② 2↑ 4↑ 4↑ 4↑ 1↑ 天气 ③ 4↓ 4↑ 5↑ 4↑ 1↑ ④ 19↓ 15↑ 13↑ 13↑ 18↑ 时间 ① 8↓ 3↑ 3↑ 3↑ 7↑ ② 1↓ 1↑ 1↑ 1↑ 1↑ 行驶状态 ④ 1↑ 1↑ ⑤ 1↓ 1↑ 1↑ 1↑ 1↓ 1↑ 1↑
下载: 导出CSV
-
[1] 宗芳, 于萍, 吴挺, 等. 常规公交风险的SEM与Bayesian Network组合评估方法研究[J]. 交通信息与安全, 2018, 36(4): 22-28. doi: 10.3963/j.issn.1674-4861.2018.04.004ZONG Fang, YU Ping, WU Ting, et al. A combination assessment of SEM and Bayesian network for safety risks of regular buses[J]. Journal of Transport Information and Safety, 2018, 36(4): 22-28. (in Chinese). doi: 10.3963/j.issn.1674-4861.2018.04.004 [2] ZHANG Yingyu, LIU Tiezhong, BAI Qingguo, et al. New systems-based method to conduct analysis of road traffic[J]. Transportation Research Part F: Traffic Psychology and Behaviour, 2018(54): 96-109. http://smartsearch.nstl.gov.cn/paper_detail.html?id=6afbbfa43f1c8d3f2afbf16954a6af13 [3] SAM E F, DANIELS S, BRIJS K, et al. Modelling public bus/minibus transport accident severity in Ghana[J]. Accident Analysis & Prevention, 2018(119): 114-121. http://www.onacademic.com/detail/journal_1000040409223510_ba95.html [4] MIYAMA G, FUKUMOTO M, KAMEGAYA R, et al. Risk factors for collisions and near-miss incidents caused by drowsy bus drivers[J] International Journal of Environmental Research and Public Health, 2020, 17(12): 1-11. [5] 陈昭明, 徐文远, 曲悠扬, 等. 基于混合Logit模型的高速公路交通事故严重程度分析[J]. 交通信息与安全, 2019, 37(3): 42-50. doi: 10.3963/j.issn.1674-4861.2019.03.006CHEN Zhaoming, XU Wenyuan, QU Youyang, et al. Severity of traffic crashes on freeways based on mixed logit model[J]. Journal of Transport Information and Safety, 2019, 37(3): 42-50. (in Chinese). doi: 10.3963/j.issn.1674-4861.2019.03.006 [6] JIANG Chenming, TAY R, LU Linjun. A skewed logistic model of two-unit bicycle-vehicle hit-and-run crashes[J]. Traffic Injury Prevention, 2021, 22(2): 158-161. doi: 10.1080/15389588.2020.1852224 [7] WANG Xuesong, JIAO Yujun, HUO Junyu, et al. Analysis of safety climate and individual factors affecting bus drivers'crash involvement using a two-level logit model[J]. Accident Analysis & Prevention, 2021(154): 1-10. http://www.sciencedirect.com/science/article/pii/S0001457521001184 [8] GULER M A, ELITOK K, BAYRAM B, et al. The influence of seat structure and passenger weight on the rollover crashworthiness of an intercity coach[J]. International Journal of Crashworthiness, 2007, 12(6): 567-580. doi: 10.1080/13588260701485297 [9] MEIRA J A D, ITURRIOZ I, WALBER M, et al. Numerical analysis of an intercity bus structure: A simple unifilar model proposal to assess frontal and semifrontal crash scenarios[J]. Latin American Journal of Solids and Structures, 2016, 13(9): 1616-1640. doi: 10.1590/1679-78252440 [10] 徐安, 乔向明. 公路客运安全分析与车辆制动性能建模[J]. 交通运输工程学报, 2009, 9(6): 87-91. doi: 10.3969/j.issn.1671-1637.2009.06.017XU An, QIAO Xiangming. Safety analysis of highway passenger transport and modeling of vehicle brake performances[J] Journal of Traffic and Transportation Engineering. 2009, 9(6): 87-91. (in Chinese). doi: 10.3969/j.issn.1671-1637.2009.06.017 [11] WANG Shuoyen, WU Kunfeng. Reducing intercity bus crashes through driver rescheduling[J]. Accident Analysis & Prevention, 2019(122): 25-35. http://www.ncbi.nlm.nih.gov/pubmed/30300796 [12] BESHARATI M M, KASHANI A T. Factors contributing to intercity commercial bus drivers'crash involvement risk[J]. Archives of Environmental & Occupational Health, 2018, 73(4): 243-250. [13] TSAI Chengjung, LEE Chieni, YANG Weipang. A discretization algorithm based on class-attribute contingency coefficient[J]. Information Sciences, 2008, 178(3): 714-731. doi: 10.1016/j.ins.2007.09.004 [14] 安宁, 藤越, 杨矫云, 等. 基于因果效应的贝叶斯网络结构学习方法[J]. 计算机应用研究, 2018, 35(12): 3609-3613. doi: 10.3969/j.issn.1001-3695.2018.12.019AN Ning, TENG Yue, YANG Jiaoyun, et al. Bayesian network structure learning method based on causal effect[J]. Application Research of Computers, 2018, 35(12): 3609-3613. (in Chinese). doi: 10.3969/j.issn.1001-3695.2018.12.019 [15] 刘飞, 张绍武, 高红艳. 基于部分互信息和贝叶斯打分函数的基因调控网络构建算法[J]. 西北工业大学学报, 2017, 35(5): 876-883. doi: 10.3969/j.issn.1000-2758.2017.05.020LIU Fei, ZHANG Shaowu, GAO Hongyan. Inferring gene regulatory networks based on part mutual information and Bayesian scoring function[J]. Journal of Northwestern Polytechnical University, 2017, 35(5): 876-883. (in Chinese). doi: 10.3969/j.issn.1000-2758.2017.05.020 [16] KRASKOV A, STOGBAUER H, GRASSBER-GER P. Estimating mutual information[J]. Physical Review E, 2004, 69(6): 1-16. http://www.ams.org/mathscinet-getitem?mr=2096503 [17] 綦小龙, 高阳, 王皓, 等. 一种可度量的贝叶斯网络结构学习方法[J]. 计算机研究与发展, 2018, 55(8): 1717-1725. https://www.cnki.com.cn/Article/CJFDTOTAL-JFYZ201808013.htmQI Xiaolong, GAO Yang, WANG hao, et al. A measurable Bayesian network structure learning method[J]. Journal of Computer Research and Development, 2018, 55(8): 1717-1725. https://www.cnki.com.cn/Article/CJFDTOTAL-JFYZ201808013.htm [18] ZHANG Yongli, YANG Yuhong. Cross-validation for selecting a model selection procedure[J]. Journal of Econometrics, 2015, 187(1): 95-112. doi: 10.1016/j.jeconom.2015.02.006 [19] 周志华. 机器学习[M]北京: 清华大学出版社, 2016.ZHOU Zhihua. Machine learning[M]. Beijing: Tsinghua University Press, 2016. (in Chinese). -
点击查看大图
图(7) / 表(6)
计量
- 文章访问数: 1365
- HTML全文浏览量: 727
- PDF下载量: 182
- 被引次数: 0