大样本与小样本巢式Logit模型标定比较  

Comparison of Nested Logit Model Calibration on Small Sample Datasets and Big Sample Datasets

作　　者：陈先龙 CHEN Xianlong(Key Laboratory of Road and Traffic Engineering of the Ministry of Education,Tongji University,Shanghai 201804,China;Information and modelling department,Guangzhou Transport Planning and Research Institute,Guangzhou 510030,China)

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201804 [2]广州市交通规划研究院信息模型所,广州510030

出　　处：《交通与运输》2020年第6期90-95,共6页Traffic & Transportation

摘　　要：为了辨识不同软件标定巢式Logit模型参数差异,选择最适宜的软件。回顾了典型交通方式划分模型巢式Logit模型结构和标定方法,基于经典小样本数据集Travelmode和大样本广州基家工作出行数据集,分别使用商业软件SAS、STATA、NLOGIT和开源软件R、Biogeme对所设计的Nested Logit模型参数进行了标定。通过对比模型参数、预测结果正确率和出行方式结构预测结果发现,对于小样本数据集,SAS和NLOGIT标定结果最优,R和STATA次之且基本一致,最后为Biogeme;对于大样本数据集,NLOGIT具有最佳的表现,其次为SAS、STATA和R,最后为Biogeme。研究表明,在Nested Logit模型标定方面,无论小样本数据集和大样本数据集,商业软件NLOGIT均是最佳选择,R则为开源软件的首选。It will help users to make appropriate choice by identifying the differences of nested logit model calibration for different software.The model structure and calibration method of the Nested Logit model were reviewed in this paper.Based on the classic small sample datasets Travelmode and the big sample Guangzhou home-based work travel datasets,the Nested Logit model parameters were calibrated by commercial software SAS,STATA,NLOGIT and open source software R,Biogeme.Comparing with model parameters,prediction accuracy and mode split result shows that SAS and NLOGIT have the best performance for small sample dataset,the secondary is R and STATA with the same result,and finally Biogeme;NLOGIT has the best performance for large sample datasets,followed by SAS,STATA and R,and finally Biogeme.This study shows that NLOGIT is the best commercial software to calibrate Nested Logit model on either small sample datasets or large sample datasets,and R is the first choice of open source software.

关 键 词：交通方式划分 巢式LOGIT模型 大样本 小样本 模型标定 开源软件 

分 类 号：U491[交通运输工程—交通运输规划与管理]