- 规划与建设
- 文章编号：1009-6000(2026)03-0065-08
- 中图分类号：TU984 文献标识码：B
- Doi：10.3969/j.issn.1009-6000.2026.03.009
- 项目基金：云南省社科规划社会智库项目“新型城镇化背景下数字乡村建设研究” （SHZK2023336）；云南省科技厅基础研究专项面上项目“多维度视角下云南省县域城镇化时空格局、动力机制及规划导控策略研究”（202301AT070408）。
- 作者简介：项振海，昆明理工大学建筑与城市规划学院，副教授；
盛杰，西南交通大学建筑学院，博士研究生；
班鹏飞，通信作者，昆明理工大学建筑与城市规划学院，特聘副教授。
- 主客观视角下的街道建成环境对青少年步行通学的非线性影响——基于极端梯度提升模型
- Nonlinear Effects of the Built Street Environment on Adolescents’ Walking to School from Objective and Subjective Perspectives: Based on the Extreme Gradient Boosting Model
- 项振海 盛杰 班鹏飞
- XIANG Zhenhai SHENG Jie BAN Pengfei
- 摘要：
以步行通学为代表的积极通学行为，对青少年体力活动的培养及身心健康的塑造具有重要意义。但目前我国关于建成环境对青少年通学行为的影响仍缺乏实证研究，同时客观环境和主观环境对步行通学的非线性作用及差异影响机制仍未被揭示。文章以昆明市为例，基于昆明市中小学生通学及健康成长问卷调查数据、百度街景数据及多源城市空间数据，采用极端梯度提升（extreme gradient boosting, XGBoost）模型，探究主客观视角下的街道建成环境对青少年步行通学的影响效应。结果表明：①客观街道建成环境对青少年步行通学的贡献显著（33.56%），远高于主观街道建成环境（15.86%）。②在单个自变量中，通学距离对步行通学的影响最为明显且相对重要性高达 30.82%，界面丰富度与通学接送情况的排序次之。③在关键主观建成环境变量中，步行通学概率与周边道路平坦程度呈非线性正相关，与周边道路行车速度则呈非线性负相关；客观建成环境变量中，界面丰富度、坡度、距地铁站最近距离与步行通学概率呈非线性正相关，交叉口密度则呈非线性负相关。 - 关键词：
建成环境；步行通学；极端梯度提升模型；昆明市；非线性 - Abstract: Encouraging active school travel behaviors, represented by walking to school, is of great significance for promoting adolescents’ physical activity and shaping their physical and mental health. However, empirical research in China on the effects of the built environment on adolescents’ school travel behavior remains limited, and the nonlinear effects and differential mechanisms of objective and perceived environments on walking to school have not yet been clearly revealed. Taking Kunming as a case study, this research uses questionnaire survey data on school travel and healthy development among primary and secondary school students, Baidu street-view data, and multi-source urban spatial data, and applies an extreme gradient boosting (XGBoost) model to examine the effects of the street built environment on adolescents’ walking to school from both objective and subjective perspectives. The results show that: (1) The objective street built environment makes a significant contribution to adolescents’ walking to school (33.56%), far higher than that of the perceived street built environment (15.86%); (2) Among individual variables, commuting distance has the most prominent effect on walking to school, with a relative importance of up to 30.82%, followed by interface richness and school travel escort conditions; (3) Among key perceived built-environment variables, surrounding road flatness shows an overall nonlinear positive relationship with the probability of walking to school, while surrounding traffic speed shows a nonlinear negative relationship; among objective built-environment variables, interface richness, slope, and distance to the nearest metro station are nonlinearly positively associated with the probability of walking to school, whereas intersection density shows a nonlinear negative association.
- Key words: built environment; walking to school; extreme gradient boosting model; Kunming; non-linearity