基于人口合成技术的居民出行调查数据扩样  

Synthetic population techniques for household travel survey data expansion

作　　者：李丁杰 乐阳 郭莉 LI Dingjie;YUE Yang;GUO Li(Shenzhen Key Laboratory of Spatial Smart Sensing and Service,Shenzhen 518060,China;Guangdong Key Laboratory of Urban Informatics,School of Architecture and Urban Planning,Shenzhen University,Shenzhen 518060,China;Shenzhen Urban Planning and Land Resource Research Center,Shenzhen 518034,China)

机构地区：[1]深圳市空间信息智能感知与服务重点实验室,广东深圳518060 [2]深圳大学建筑与城市规划学院广东省城市空间信息工程重点实验室,广东深圳518060 [3]深圳市规划国土发展研究中心,广东深圳518034

出　　处：《交通科技与经济》2021年第6期24-31,共8页Technology & Economy in Areas of Communications

基　　金：国家自然科学基金面上项目(41671387)。

摘　　要：为提高居民出行入户调查样本代表性,使调查数据能够准确反映居民出行特征,在分析居民出行调查数据处理现状及常规扩样方法局限性的基础上,引入基于人口合成技术的IPU算法(iterative proportional updating)和GRE算法(generalized regression),对深圳市居民出行入户调查数据进行实证分析,并利用相关系数及相对误差指标对扩样结果进行评价。结果表明:利用算法关联家庭与个人表的频数矩阵,通过迭代调整扩样权重值、IPU算法和GRE算法均能很好地拟合家庭与个人的属性分布与总体一致,且相对误差控制在6%左右,满足扩样的精度要求。由于IPU算法更具普适性,算法不受初始权重及稀疏样本限制,其扩样误差及波动性较小,扩样结果更为稳健。In order to strengthen the sample of household travel survey,the result can reflect the actual travel characteristics.On the analysis of resident trip survey data processing status and limitations of conventional expansion method,this paper introduces the synthetic population techniques which use both IPU and GRE algorithms to fill up the research gap.Taking Shenzhen residents travel survey data as a case,and using correlation coefficient and relative error to evaluate the expansion results,the result shows that through iterative calculations,IPU and GRE algorithms that correlate the frequency matrix of the household and individual tables,perform well,and the relative error is about 6%.These algorithms meet the requirements in accuracy.Moreover,IPU algorithm is more universal,not limited by initial weight and sparse samples,and the error and fluctuation of the IPU algorithm are smaller and more robust.

关 键 词：居民出行调查 数据扩样 人口合成技术 IPF算法 IPU算法 GRE算法 

分 类 号：U491[交通运输工程—交通运输规划与管理]