基于Adaptive Lasso及RF算法的冰雪天气交通事故分析

发布时间：2018-05-31 15:27

本文选题：高速公路 + 交通事故　；参考：《中国安全科学学报》2017年02期

【摘要】：为分析冰雪天气下高速公路交通事故频发致因,量化分析驾驶环境、驾驶员及车辆情况对事故的影响,根据Adaptive Lasso和随机森林(RF)混合算法建立预测模型。以10年约30万组冰雪环境下高速公路交通事故数据为例,训练改进预测模型验证其准确性。结果表明,混合算法的准确度和拟合程度都优于支持向量机(SVM)、分类回归树(CART)及RF等单独算法。交通事故与环境因素相关性最显著,坡路、弯道及交叉口处事故受冰雪环境影响较大;事故与驾驶员因素中部分因素显著相关,如驾驶员性别及安全带使用情况;本地驾驶员对驾驶能力及冰雪环境的估计错误更易导致交通事故。
[Abstract]:In order to analyze the frequent causes of traffic accidents in freeway under ice and snow weather, and to quantitatively analyze the influence of driving environment, driver and vehicle situation on the accidents, a prediction model was established based on the hybrid algorithm of Adaptive Lasso and random forest fasts. Taking about 300000 groups of highway traffic accident data under ice and snow environment for 10 years as an example, the improved prediction model is trained to verify its accuracy. The results show that the accuracy and fitting degree of the hybrid algorithm are better than the single algorithms such as support vector machine (SVM), classification regression tree (CARTT) and RF. Traffic accident and environmental factors are the most significant, slope road, bend and intersection accidents are greatly affected by ice and snow environment, accidents and driver factors are significantly related to some factors, such as driver gender and seatbelt use. Local drivers' estimation of driving capacity and ice-snow environment is more likely to lead to traffic accidents.
【作者单位】：武汉理工大学交通学院;内蒙古科技大学经管学院;威斯康星州立大学工程学院;东南大学交通学院;
【基金】：教育部社会科学青年基金资助(16YJCZH157) 内蒙古科技大学创新基金资助(2015QDL27) 国家重点基础研究发展(“973”)计划项目(2012CB725405)
【分类号】：U491.31

【参考文献】