大数据时代交通需求模型范式转变的思考  

Reflections on the Paradigm Shift in Travel Demand Models in the Big Data Era

作　　者：陈先龙 张华 马毅林 宋程 魏贺[5] CHEN Xianlong;ZHANG Hua;MAYilin;SONG Cheng;WEI He(Guangzhou Transport Planning Research Institute Co.,Ltd.,Guangzhou Guangdong 510030,China;Guangdong Sustainable Transportation Engineering and Technology Research Center,Guangzhou Guangdong 510030,China;National Maglev Transportation Engineering R&D Center,Tongji University,Shanghai 201804,China;Beijing Transport Institute,Beijing 100073,China;Beijing Municipal Institute of City Planning&Design,Beijing 100045,China)

机构地区：[1]广州市交通规划研究院有限公司,广东广州510030 [2]广东省可持续交通工程技术研究中心,广东广州510030 [3]同济大学磁浮交通工程技术研究中心,上海201804 [4]北京交通发展研究院,北京100073 [5]北京市城市规划设计研究院,北京100045

出　　处：《城市交通》2024年第4期50-57,105,共9页Urban Transport of China

基　　金：国家自然科学基金项目“双边-配对市场视角下类出租车服务平台算法治理研究”(72174147);广州市交通规划研究院有限公司科研项目“数据驱动的时空推演城市活动模型研究”(KYHT-2023-01)。

摘　　要：大数据给交通需求模型开发基础带来了革命性的变化。对范式转变的思考既是对新时代数据条件变化的适应,也是提升交通需求模型精度的必然要求。基于基础数据条件的变化,对交通需求模型四种范式转变进行总结和思考:从数学优化到因果推断,强化模型对出行行为的解释能力;从比例因子到概率抽样,使得模型的物理意义更为明确;从整体重构到增量模型,实现对现状需求的继承和迭代演化;从有限约束收敛到先验实证,提升交通需求模型精度。指出对于交通治理实践运用和科学研究而言,模型精度是检验模型质量的唯一和最高标准。对范式转变的重视不是为了否定既往交通需求模型技术路线,而是在继承中创新发展,提高交通需求模型对现实物理世界的模拟能力和预测能力。Big data has brought revolutionary changes to the developing basis of travel demand models.Reflections on the paradigm shift is not only an adaptation to the changes in data conditions in the new era,but also a necessary requirement for improving the accuracy of travel demand models.This paper summarizes and contemplates four paradigm shifts in travel demand models based on changes in foundational data conditions.These shifts include:enhancing the explanatory power of models on travel behavior by transitioning from mathematical optimization to causal inference;clarifying the physical meaning of models by moving from proportional factors to probabilistic sampling;achieving inheritance and iterative evolution of current demand by transitioning from holistic reconstruction to incremental models;and improving the precision of travel demand models by moving from finite constraint convergence to prior empirical justification.The paper points out that,for both practical application and scientific research in transportation governance,model accuracy is the sole and highest standard for evaluating model quality.Emphasizing paradigm shifts is not meant to negate previous technical approaches in travel demand models but to innovate and develop through inheritance,thus enhancing the capacity of models to simulate and predict the real world.

关 键 词：交通需求模型 大数据 因果推断 概率抽样 增量模型 先验实证 范式转变 

分 类 号：U491.12[交通运输工程—交通运输规划与管理]