基于选择方案抽样调查的城市群低频率出行行为研究  

Study on Low-frequency Intercity Travel Behavior of Urban Agglomeration Based on Choice-based Sampling Survey

作　　者：陈颖雪[1,2] 董治[1] 吴兵[1] 刘志钢[2] 

机构地区：[1]同济大学道路与交通工程教育部重点实验室,上海201840 [2]上海工程技术大学城市轨道交通学院,上海201620

出　　处：《中国公路学报》2013年第3期158-163,共6页China Journal of Highway and Transport

基　　金：国家自然科学基金项目(51178346);国家自然科学基金青年科学基金项目(70901057);上海市科委地方院校计划能力建设项目(11170501400)

摘　　要：为分析城市群内低频率城际出行行为,对长三角城市群内上海与其他城市间的城际出行进行选择方案抽样调查,基于城际出行行为调查数据建立出行方式选择的多项Logit(MNL)模型;选择方案抽样调查数据与随机抽样调查数据存在方案分层结构偏差和因素重视度偏差,故引入反转因子(IPW),分别用外源性样本极大似然估计(ESMLE)简单修正法和加权外源性样本极大似然估计(WESMLE)法对MNL模型进行修正。结果表明:未修正模型及2种修正法下的模型命中率分别为57%,75%和84%;WESMLE修正后的模型能基本满足建模精度要求;城市群内低频率城际出行行为受出行时间、费用及个人属性等因素影响,同时,也表现出一定程度的随机性和随意性。In order to analyze low-frequency intercity travel behavior in urban agglomeration, a questionnaire survey with choice-based sampling was made on intercity travel between Shanghai and the other cities in Yangtze River delta, then multinomial Logit （MNL） model was built based on intercity travel behavior data collected. Since data from choice-based sampling were of endogenous stratification bias and avidity bias compared with data from random sampling, inverse-probability weights （IPWs） were introduced and simple modification on exogenous sample maximum likelihood estimator （ESMLE） and weighted exogenous sample maximum likelihood estimator （WESMLE） were used to modify MNL model. The results show that the hit rate of the model without modification, with ESMLE simple modification and WESMLE modification is 57%, 75% and 84% respectively; MNL model with WESMLE modification can meet the required accuracy; low-frequency intercity travel behavior is influenced by travel time, cost and personal characteristics, meantime, the behavior is of some randomness and haphazardness.

关 键 词：交通工程 城际出行行为 选择方案抽样法 MNL 低频率 反转因子 

分 类 号：U491.1[交通运输工程—交通运输规划与管理]