文章编号： 1672-5328（2025）03-0074-11

陈先龙1, 2，马毅林3，宋程1, 2，陈嘉超1, 2
(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.北京交通发展研究院北京100073)

摘要： 为验证居民出行调查加权扩样模型的效果，从已知总体的先验视角研究样本代表性和扩样结 果可靠性。构建了由居民出行调查样本组成的小型总体数据集和基于活动模型出行活动仿真结果的 大型总体数据集，设计并搭建可重复验证的实验平台，模拟居民出行调查抽样和扩样过程并进行实 证研究。结果表明：抽样率对出行距离分布存在影响，抽样率越低，平均出行距离波动越显著，同 时揭示了加权扩样模型扩样系数的分布规律。仿真实验发现：加权扩样会降低平均出行距离，推测 其原因为出行距离呈左偏分布，并基于均匀分布和正态分布构造出行距离数据集验证了这一猜想。 实证分析显示：针对小型总体和大型总体数据集，即便抽样率达30%，扩样加权方法仍无法还原高 质量的出行OD矩阵。最后指出，居民出行调查本质是小样本特征调查，可作为交通模型构建与校 验的基础数据，其扩样结果不宜直接用于城市出行活动空间分布分析。

关键词： 居民出行调查；抽样；加权扩样；仿真实验；先验视角

中图分类号： U491.1

文献标识码：A

A Simulation Study on Sampling and Expansion of Household Travel Survey from a Prior Perspective

CHEN Xianlong1, 2, MAYilin3, SONG Chen1, 2, CHEN Jiachao1, 2
(1. Guangzhou Transport Planning Research Institute Co., Ltd., Guangzhou Guangdong 510030, China; 2. Guangdong Sustainable Transportation Engineering and Technology Research Center, Guangzhou Guangdong 510030, China; 3. Beijing Transport Institute, Beijing 100073, China)

Abstract： From a priori perspective with known population datasets, this paper demonstrates the effectiveness of weighted expansion models used in household travel survey by assessing sample representativeness and reliability of expansion result. A small size population dataset is constructed from survey samples and a large size population dataset is generated through activity-based model. A reproducible experimental platform is then developed to simulate the full sampling and expansion process in an empirical study. Findings indicate that sampling rates affect trip length distribution: lower sampling rates are associated with more fluctuation in average travel distance. The simulations also reveal the distribution pattern of the model's expansion factors. According to simulation experiments, weighted expansion tends to reduce the average travel distance likely due to its left-skewed distribution; this hypothesis is validated using synthetic datasets created under uniform and normal distributions. Empirical analysis further demonstrates that, even at a 30% sampling rate, the weighted expansion method fails to reconstruct a high- quality origin- destination (OD) matrix for both small- and large-size population datasets. Finally, the paper indicates that household travel survey is fundamentally small size sample attribute survey and can serve as foundational data for constructing and calibrating transportation models; however, the expanded results should not be directly used to analyze the spatial distribution of urban travel activities.

Keywords： household travel survey; sampling; weighted expansion; simulation experiment; prior perspective