【作者】 王慧；

【导师】 吴茜茜；

【作者基本信息】 合肥工业大学 ， 概率论与数理统计， 2020， 硕士

【摘要】 自改革开放以来,随着我国大力发展经济和推进城市化水平,交通拥堵及环境污染问题日益严重,人们出行效率低下,从而阻碍城市和经济发展。为了讨论出行方式选择问题,本文以认知心理学和行为理论为框架,引入出行者环境偏好心理这一潜在变量构建结构方程模型和离散选择模型相结合的混合选择模型,并利用贝叶斯方法对模型中的未知参数进行估计,从而对出行者出行行为进行理论的分析和应用的研究。本文具体内容如下:第一章介绍出行方式选择问题的研究背景及现状;第二章介绍相关理论基础,其中主要包含贝叶斯定理、马尔科夫链蒙特卡洛方法(Metropolis-Hasting和Gibbs抽样方法)、近似贝叶斯计算的基本思想和拒绝算法;第三章为模型的简介和构建,其中包括通过采用Likert5级量表设计出行方式问卷收集样本,基于潜山市样本数据的具体信息引入环境偏好潜在变量构建混合选择模型,并提出应用于该模型的马尔科夫链蒙特卡洛算法和近似贝叶斯算法;第四章基于提出的算法利用数据对模型进行参数估计,分析结果并将两种方法进行比较;第五章为总结及展望。实证结果表明环境偏好心理这一潜在变量对居民出行方式选择意向有影响;女性、18-40岁之间、大学及以上的人会更加注重环境因素,其选择WB出行方式的可能性更高;出行方式的碳排量、方便程度是影响居民出行考虑的重要因素。此外,基于马尔科夫链蒙特卡洛方法虽依赖于似然的核函数,对于先验超参数的设定更为敏感,且要求控制采样样本的自相关性,却可以高效地估计参数;而近似贝叶斯计算能够避免对似然函数的求解,以及对于先验超参数的设定更稳固,同时也较为有效地估计参数。更多还原

【Abstract】 Since the reform and opening-up,Chinese economy has developed rapidly while the level of urbanization has improved,which have brought traffic congestion and serious environmental pollution problems,leading to low efficiency of people travel,hindering urban and economic development.Aiming at studying the personal choice of travel mode,a hybrid choice model combining structural equation model and measurement equation model is constructed by introducing traveler’s psychology as potential variables based on cognitive psychology and behavioral theory,and Bayesian methods are used for estimating unknown parameters in the model to discuss theoretical and applicable aspects of traveling behaviors.This paper is carried out from the following aspects: the first chapter introduces the research background and current research status of the choice of travel mode.The second chapter introduces the theory,including Bayes’ theorem,Markov Chain Monte Carlo method(Metropolis-Hasting and Gibbs sampling method),the basic idea of approximate Bayes computation,and the rejection algorithm.The third chapter presents an introduction and construction of the model,including the questionnaire collection of samples designed by likert5-level scale;the hybrid choice model constructed by introducing the latent variable of environmental preferences based on the data collected for Qianshan city;and the Markov Chain Monte Carlo algorithm and approximate Bayesian algorithm applied to the model.In chapter 4,the unknown parameters of the model are estimated based on the proposed algorithm with data collected,the results are analyzed and the two methods are compared.The last chapter summarizes the work and provides future directions.The empirical results show that the psychological potential variables of environmental preference have an impact on the residents’ travel intention.Female,18-40 years old,and those people who has a bachelor degree or above will pay more attention to environmental factors,and are more likely to choose WB travel mode.The carbon emission and convenience of the travel mode are the main factors affecting the residents’ travel consideration.By comparing the two algorithms,the Markov Chain Monte Carlo method which is dependent on the likelihood of kernel function,is found to be more sensitive to priors’ hyper-parameter values and requires controlling sample’s correlation,however,it can estimate parameters highly efficiently.The approximate Bayesian computation which can avoid solving the likelihood function,is more robust to the values of the prior hyper-parameters and can estimate the parameters effectively.更多还原