Scholarlink

<p>Rapid urban renewal presents both opportunities and challenges for sustainable mobility in high-density cities, particularly by intensifying traffic congestion. However, the complex interactions between land use transformation and traffic performance remain insufficiently understood, particularly at fine spatial scales. This study develops an explainable machine-learning framework to forecast traffic congestion associated with urban renewal and to elucidate how the building capacity of different functional transformations across multiple spatial scales relates to traffic performance. Using Shenzhen’s Longgang District as a case study, we integrate high-resolution land use data from multiple buffer zones with socio-economic, temporal, and road network characteristics. The framework leverages XGBoost for prediction and SHAP for model interpretation, enabling the identification of nonlinear relationships and development-intensity thresholds associated with urban land transformation. The results show that moderate residential and mixed-use capacities can alleviate congestion, whereas high-intensity redevelopment, particularly when unsupported by adequate infrastructure, is strongly associated with heightened traffic congestion. These predictive variations vary with proximity range, land use type, and location. By incorporating fine-grained, function-specific intensity thresholds, this approach extends traditional land use-transport interaction models and provides a capacity-informed, location-sensitive basis for urban renewal planning. The proposed workflow offers planners a transparent and data-driven tool to anticipate and manage the traffic outcomes of urban renewal.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Forecasting Traffic Congestion Caused by Urban Renewal: Analysing the Predictive Influence of Land Use Changes Based on Explainable Machine Learning Methods

  • 期刊论文

  • 发表日期：

    2026年7月16日

查看全文
  • He Zhu,
  • Meiyu Wei,
  • Liu Yang,
  • Tianping Zhang

摘要

Rapid urban renewal presents both opportunities and challenges for sustainable mobility in high-density cities, particularly by intensifying traffic congestion. However, the complex interactions between land use transformation and traffic performance remain insufficiently understood, particularly at fine spatial scales. This study develops an explainable machine-learning framework to forecast traffic congestion associated with urban renewal and to elucidate how the building capacity of different functional transformations across multiple spatial scales relates to traffic performance. Using Shenzhen’s Longgang District as a case study, we integrate high-resolution land use data from multiple buffer zones with socio-economic, temporal, and road network characteristics. The framework leverages XGBoost for prediction and SHAP for model interpretation, enabling the identification of nonlinear relationships and development-intensity thresholds associated with urban land transformation. The results show that moderate residential and mixed-use capacities can alleviate congestion, whereas high-intensity redevelopment, particularly when unsupported by adequate infrastructure, is strongly associated with heightened traffic congestion. These predictive variations vary with proximity range, land use type, and location. By incorporating fine-grained, function-specific intensity thresholds, this approach extends traditional land use-transport interaction models and provides a capacity-informed, location-sensitive basis for urban renewal planning. The proposed workflow offers planners a transparent and data-driven tool to anticipate and manage the traffic outcomes of urban renewal.