站域建成环境与地铁客流量的非线性关系和协同效应——可解释机器学习分析
Nonlinear and Synergistic Effects of Station-area Built Environments on Metro Ridership:A Shapley Additive Explanations(SHAP)Analysis
汪雨菲 1杨皓森 2喻冰洁 2付飞 2杨林川2
作者信息
- 1. 电子科技大学经济与管理学院,成都 611731
- 2. 西南交通大学建筑学院,成都 611756
- 折叠
摘要
为了量化站域建成环境与地铁客流量的复杂关联效应,运用公共交通刷卡、手机信令、POI(point of interest)等多源大数据,采用可解释机器学习方法(融合随机森林和SHAP(Shapley additive explanations)模型),对站域建成环境变量与成都地铁客流量之间的非线性关系以及变量之间的协同效应进行实证研究.研究结果表明:对地铁客流量影响最大的3个建成环境变量是容积率、就业密度和道路密度.SHAP模型分析进一步揭示了站域建成环境对地铁客流量的阈值效应以及建成环境变量之间的协同效应.上述发现为以公共交通为导向的城市发展(transit-oriented development,TOD)规划和实践提供了理论支持和政策启示.
Abstract
This study uses multi-source big data(e.g.,metro card transactions,mobile phone signaling,and points of interest(POIs))and interpretable machine learning methods(integrating random forest and Shapley additive explanations(SHAP)models)to investigate the nonlinear relationship between station-area built environments and Chengdu Metro ridership as well as the synergistic effects among built environment variables.The results indicate that the three most important built environment determinants of metro ridership are the floor area ratio,employment density,and road density.Moreover,the SHAP model results reveal the threshold and synergistic effects of the station-area built environment variables on metro ridership.These findings provide theoretical support and policy insights for transit-oriented developmental(TOD)planning and practice.
关键词
城市轨道交通/TOD/建成环境/协同效应/非线性/机器学习Key words
urban rail transit/transit-oriented development(TOD)/built environment/synergistic effect/nonlinearity/machine learning引用本文复制引用
出版年
2024
NETL
NSTL
万方数据