城市交通出行行为潜变量作用机理及整合模型研究

The existing models of urban travel behavior analysis considered only the influence of manifest variables (MV) which can be directly observed, but did not systematically consider the role of latent variables (LV) in external environment and internal of travelers. Therefore the expression ability and accuracy of existing models were questioned, and it was difficult to quantitatively describe the psychological decision-making mechanism. From the perspective of social psychology, this project was conducted based on the models and methods of TPB, EMB, SEM, ML/NL, MSL and etc. The main contents include: 1. The influence mechanism of latent variables in the "black box" of travel behavior was studied, and then the law of choice difference in the same decision process that made by travelers with different personality traits was analyzed. 2. The integrative model affected by both LV and MV was built. The interactive quantitative relationship between the different types of LV, between the same type of LV, and between different types of LV and their corresponding measurement variables in the same group was revealed. The different degree of path coefficient of the same relationship among different groups was revealed. The sensitivity of different types of LV which affect travel choices result of integrative model was explored. 3. An integrative algorithm of LV measurement and parameters calibration of integrative model was designed. And the recommended values of the optimal simulation times for difference sample sizes were proposed. With the concept of skewness, the effect of different measurement methods for the same LV was calculated. The research findings could provide a more accurate analysis method for urban transportation planning and management, and it would enrich and develop the travel behavior theory.

现有城市交通出行行为分析模型中仅考虑可直接观测的显变量影响，尚未系统研究外界与出行者内在潜变量的作用，使得模型的解释能力和精度受到质疑，难以定量描述出行行为心理决策机理。本项目从社会心理学的视角，基于TPB、EMB、SEM、ML/NL、MSL等模型方法，开展：1、交通出行行为"黑盒子"中潜变量的作用机理研究，分析不同人格特质的出行者在相同决策过程中的选择差异规律；2、构建潜变量与显变量共同作用的整合模型，揭示群体内不同类型潜变量间、相同类型潜变量间、各潜变量与其对应测量变量间相互作用定量关系及不同群体间相同关系路径系数差异程度，探索各潜变量对整合模型结果影响的灵敏度；3、设计潜变量测度与整合模型标定一体化算法，提出各样本量的最优模拟次数推荐值，引入偏度的概念用以计算潜变量不同测度方式的效果差异。研究成果将完善与丰富交通出行行为理论，为城市交通规划、管理等提供更准确的分析方法。

城市交通出行行为是交通需求预测、交通规划与管理等的出发点与前提，本课题以出行行为中的潜变量为研究对象，系统开展了潜变量的作用机理、涵盖潜变量的整合模型、一体化算法及实证分析等研究。.首先，从心理学的视角解析出行行为本质，分析已有出行行为建模方法的特点，提出潜变量研究的必要性，对出行行为中潜变量内涵、应用领域、模型构建进行系统梳理。根据潜变量描述对象的差异将出行行为研究中到的潜变量分为情景性潜变量、个人感知性潜变量、无形服务性潜变量三大类，形成了潜变量图谱。.其次，基于计划行为理论构建出行行为个人感知性潜变量间关系模型，运用结构方程模型分析方法对模型进行适配度检验，通过变量间的直接或间接影响关系以及不同群体的行为差异，解构潜变量对出行选择行为结果的影响机理。运用成都、重庆3次大规模问卷调查所获得的数据进行实证分析，结果表明：知觉行为控制、出行者态度、主观规范3个潜变量对出行行为意向的解释度可达到60-70%。同时，在EMB（Engel Miniard Blackwell model）架构的基础上研究城市公交方式选择行为中知觉价值、服务品质、价格合理性等潜变量相互因果关系及对选择结果的影响大小，提出了一种新的方式选择行为分析方法。.然后，通过结构方程模型刻画方式选择行为中潜变量与显变量、潜变量与其测量变量之间的因果关系，掌握潜变量的适配值。基于个人离散选择模型（DCM），对出行方式效用函数进行改进，构建了潜变量与显变量共同作用的SEM-DCM整合模型。并采用最大似然估计法推导SEM与DCM一体化算法，实证结果证明整合模型较传统Logit模型解释能力和精度有了较大提高。.最后，探索了城市交通出行方式对区域出行方式选择行为的影响，在分析区域出行方式特性影响因素的基础上，考虑城市内部出行的特征因素，确定出行全过程影响指标，构建了含有城市内部出行影响因素Nested Logit模型。本课题还拓展研究了将城市交通出行行为中的潜变量转化为出行路径中的潜费用，增加至各路径的总效用中，建立了考虑潜费用的多层次、多类型弹性需求随机用户平衡配流模型。.课题研究成果将弥补现有出行行为分析方法与模型仅考虑可直接观测的显变量影响，尚未系统研究出行者内在潜变量作用的不足，丰富出行行为领域的理论成果，为更加精细化的出行需求分析、交通战略制定提供决策支持，具有重要的实践指导意义。