Computer Engineering and Applications ›› 2015, Vol. 51 ›› Issue (13): 1-8.

Previous Articles     Next Articles

Advanced discrete choice Logit models and estimation and forecast: Case in large scale residential area on Shanghai city periphery

GUAN Jinping1,2, YANG Dongyuan1   

  1. 1.The Key Laboratory of Road and Traffic Engineering, Ministry of Education, School of Transportation Engineering, Tongji University, Shanghai 200092, China
    2.Department of Civil and Environmental Engineering, Institution of Transportation Studies, University of California, Berkeley, USA
  • Online:2015-07-01 Published:2015-06-30

离散选择高阶段Logit模型的建模与预测
——以上海城市外围大型居住社区为例

关金平1,2,杨东援1   

  1. 1.同济大学 交通运输工程学院 道路与交通工程教育部重点实验室,上海 200092
    2.美国加州大学伯克利分校 土木与环境工程系 交通研究中心,美国

Abstract: The series of discrete choice Logit models, from basic to advanced are: binominal, multinomial, nested, cross nested, mixed Logit model, latent class model and hybrid choice model with latent variables and dynamic Logit model. Developing and using advanced Logit models is the latest key technique when the residents’ mode choice is studied in the rapid urbanization process. As a professor at University of California, Berkeley, Daniel L McFadden contributes to the discrete choice theory and method and won the Nobel Economic Prize in 2000. From year 2012 to 2014, the author worked with Joan Walker and Kenneth Train seized the opportunity of learning the series of Logit models in-depth. With theory applying to practice, this paper presents the advanced Logit models and the results in the case of the large scale residential area on Shanghai city periphery in the period of Chinese “the Twelfth Five-Year Plan” and the process of rapid urbanization. With the data from the residents in the large scale residential area on Shanghai city periphery, an advanced Logit mode choice model’s estimation and forecast are developed. Results show that in the six policy scenarios, building retails closer improves the residents’ travel quality the most and supports transit priority. Combining theory with practice, advanced Logit model can support policy decision making process quantitatively and excellently.

Key words: advanced discrete choice Logit models, rapid urbanization process, the large scale residential area on Shanghai city periphery, residents&rsquo, mode choice model estimation and forecast, support decision making process quantitatively

摘要: 离散选择Logit系列模型的发展层次顺序是:二项、多项、巢式与交叉巢式、混合Logit(连续及离散潜在种类模型)、嵌入潜在变量的混合选择模型、动态Logit模型。建立高阶段Logit模型算法是当前研究快速城市化进程中居民交通方式选择与预测的最新技术手段。2000年,Daniel M教授,以“对分析离散选择的原理和方法所做出的发展和贡献”获诺贝尔经济学奖。2012—2014年,作者深入学习了离散选择Logit系列模型及标定预测,结合上海城市外围大型居住社区在中国“十二·五”期间,快速城市化进程中的演变,开展了基于高阶段Logit建模算法研究,取得了理想的成果。针对高阶段Logit模型(混合Logit模型、混合选择模型、动态Logit模型)进行了建模、算法、选择概率与消费者剩余预测的理论分析,建立了上海市外围大型居住社区居民交通方式高阶段选择模型。结果显示,6种政策方案中,附近设商业能最大改善出行质量并保证公交优先。高阶段Logit模型理论实际结合后,能很好地发挥其为政策方案决策定量支持之作用。

关键词: 离散选择高阶段Logit模型, 快速城市化进程, 上海市外围大型居住社区, 居民交通方式选择建模与预测, 可量化的决策支持