基于CART改进模型的驾驶员路径选择行为估计方法

发布时间：2018-10-25 07:45

【摘要】：针对目前驾驶员路径选择估计精度不高的问题,考虑地区、城镇类型、性别、年龄、是否已婚、学历、职业、是否从事全职工作、收入水平、交通拥挤程度、排队长度、延误时间、道路熟悉程度、路径长度、替代路径节省时间等多方面因素,设计了一个驾驶员路径选择行为调查方案,并对驾驶员群体开展了网上调查.利用Logit模型及Probit模型分析了驾驶员路径选择行为的影响因素,得到性别、年龄、是否从事全职工作、延误时间、道路熟悉程度、路径长度、道路拥挤程度等因素的影响显著.利用改进的分类树(classification and regression tree,CART)模型设计了驾驶员路径选择行为估计模型,重点针对驾驶员路径选择行为的特点对传统CART模型的递归划分与剪枝2个主要算法进行了改进研究.样本测试结果表明:模型的估计精度可达82%,相比现有模型的估计精度至少提高了6%.研究成果可为交通诱导方案的制定提供有效的技术支持.
[Abstract]:In view of the problem that the estimation accuracy of driver's path selection is not high at present, consider the area, town type, sex, age, married, education, occupation, whether engaged in full-time work, income level, traffic congestion degree, queue length, Based on the factors of delay time, road familiarity, path length and alternative route saving time, a route selection behavior investigation scheme is designed, and an online investigation of driver group is carried out. By using Logit model and Probit model, the influence factors of driver's path choice behavior were analyzed. The factors such as gender, age, whether engaged in full-time work, delay time, road familiarity, path length, road congestion were significantly affected. Based on the improved classification tree (classification and regression tree,CART) model, the estimation model of driver's path selection behavior is designed, and the two main algorithms of recursive partition and pruning of traditional CART model are studied in detail according to the characteristics of driver's path selection behavior. The sample test results show that the estimation accuracy of the model can reach 82%, which is at least 6 times higher than that of the existing model. The research results can provide effective technical support for the formulation of traffic guidance schemes.
【作者单位】：中国刑事警察学院交通事故处理教研室;公安部痕迹检验鉴定技术重点实验室;沈阳农业大学工程学院;宁波大学海运学院;国家道路交通管理工程技术研究中心宁波大学分中心;
【基金】：国家自然科学基金资助项目(51305041) 公安部技术研究计划面上项目(2015JSYJC06) 辽宁省自然科学基金资助项目(201602809) 辽宁省教育厅科学技术研究一般项目(L2015554)
【分类号】：U491

【参考文献】