Scholarlink

<p>This study presents a user-centric machine learning (ML) framework for predicting multi-modal accessibility in Transit-Oriented Development (TOD) zones across tier-2 Indian cities. Using survey data from 400 respondents in Bhopal, Jaipur, Nagpur, and Coimbatore, the research integrates socio-demographic, behavioral, and perceptual variables to model accessibility scores that reflect real-world user experiences. Seven supervised ML models—Random Forest, Support Vector Regressor, Artificial Neural Network, Decision Tree, XGBoost, CatBoost, and Extra Trees—were evaluated using R², RMSE, and MAE. XGBoost emerged as the top performer with an R² of 0.879. SHAP analysis revealed integration with public transport, physical access, affordability, safety, and user satisfaction as the most influential features. Sensitivity analysis and Partial Dependence Plots validated the robustness and policy relevance of the top predictors. Further, disaggregated analysis by gender, age, and income highlighted critical equity gaps in perceived accessibility. The framework aligns with SDG 11.2 by providing a scalable decision-support tool to identify intervention points for inclusive and sustainable TOD planning. The study offers actionable insights for urban policymakers, transport planners, and smart city missions, emphasizing the integration of subjective user experience with data-driven predictive modeling to promote equitable urban mobility outcomes.</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

A user-centric machine learning framework for predicting multi-modal accessibility in transit-oriented development zones for sustainable urban construction in tier-2 Indian cities

  • 期刊论文

  • 发表日期：

    2026年3月9日

查看全文
  • Krishna Yadav,
  • Kavita Dehalwar,
  • Shashikant Nishant Sharma

摘要

This study presents a user-centric machine learning (ML) framework for predicting multi-modal accessibility in Transit-Oriented Development (TOD) zones across tier-2 Indian cities. Using survey data from 400 respondents in Bhopal, Jaipur, Nagpur, and Coimbatore, the research integrates socio-demographic, behavioral, and perceptual variables to model accessibility scores that reflect real-world user experiences. Seven supervised ML models—Random Forest, Support Vector Regressor, Artificial Neural Network, Decision Tree, XGBoost, CatBoost, and Extra Trees—were evaluated using R², RMSE, and MAE. XGBoost emerged as the top performer with an R² of 0.879. SHAP analysis revealed integration with public transport, physical access, affordability, safety, and user satisfaction as the most influential features. Sensitivity analysis and Partial Dependence Plots validated the robustness and policy relevance of the top predictors. Further, disaggregated analysis by gender, age, and income highlighted critical equity gaps in perceived accessibility. The framework aligns with SDG 11.2 by providing a scalable decision-support tool to identify intervention points for inclusive and sustainable TOD planning. The study offers actionable insights for urban policymakers, transport planners, and smart city missions, emphasizing the integration of subjective user experience with data-driven predictive modeling to promote equitable urban mobility outcomes.