摘要
山区公路复杂的组合驾驶环境因素与事故数据的缺乏记录都使得分析山区公路驾驶安全性十分困难。建立了一套完整的山区公路驾驶安全性实验方法,并提供了一种结合了随机森林法则与基于机器学习决策树的方法,以分析山区公路上导致驾驶事故的主要因素。实验团队使用行车记录仪结合视频识别技术以获取沿途的道路驾驶环境与驾驶员在试验中的驾驶行为。用"最大信息熵增长率"法结合AIC准则以分析归纳模型中包含的主要驾驶环境因素。结论显示:急窄弯道、穿行于村镇间、缺乏视距是影响驾驶员在山区公路上安全行驶的主要原因。证明了这种方法可以用于指导山区公路的设计,提高山区公路上的行驶安全性。
The combination of complex driving environment factors and the lack of accident data make the analysis of driving safety on mountain highway difficult. This paper proposed a complete set of experimental methods for driving safety on mountain highway and provided a decision tree method combining with Random Forest method to identify the main factors leading to accidents. By using automobile data recorder and radio frequency identification, the experimental team obtained information about the driving environment and drivers＇ behaviors.The approach of maximum gain ratio and Akaike information criterion are adopted to analyze and summarize the main driving environment factors involved in the model. The results showed that sharp road curvatures, traveling through villages and lack of sight distance are the main factors for driving safety on mountain highway. It is confirmed that this approach can provide reference for the design of mountain highway to improve driving safety.
作者
李卓
陈雨人
Li Zhuo Chen Yuren(College of Transportation Engineering,Tongji University ,Shanghai 201804 ,China)
出处
《华东交通大学学报》
2017年第2期29-36,共8页
Journal of East China Jiaotong University
基金
国家科技支撑计划课题(2014BAG01B06)
关键词
驾驶安全性实验
驾驶环境因素
决策树分类
driving safety experiment
driving environment factors
decision tree method
