Scholarlink

<p>Urban growth modelling is crucial for sustainable development, especially in rapidly urbanizing regions where balancing infrastructure expansion with environmental sustainability poses ongoing challenges. Traditional machine learning (ML) models often fall short in capturing the complex, non-linear patterns of urban systems. To address this, we propose a hybrid intelligence framework that integrates linear, tree-based, and neural network models to enhance predictive accuracy and interpretability. Using geospatial and socio-economic datasets for Lucknow City, India, we developed three novel hybrid models: Logistic Regression-XGBoost-Feedforward Neural Network (LR-XGB-FNN), ElasticNet-LightGBM-TabNet (EN-LGBM-TN), and Generalized Linear Model-CatBoost-Wide &amp; Deep Network (GLM-CB-WDN). These models were trained on urban growth patterns between 2011 and&#xa0;2021, derived from Landsat imagery and thematic indices, and validated through change detection analysis. Among them, LR-XGB-FNN achieved the highest accuracy (94.17%) and AUC (0.9475). Spatial probability maps were generated to forecast urban expansion hotspots for 2031, offering actionable insights for planners. To improve transparency, SHAP (SHapley Additive exPlanations) was used to interpret model outputs, revealing Drive Time (DT), Elevation, and Distance from Roads (DfR) as key drivers. This study contributes to the field by (1) proposing robust hybrid ML models, (2) applying explainable AI for interpretability, (3) producing future urban growth maps, and (4) presenting a scalable framework aligned with SDG 11. The results demonstrate how hybrid intelligence can support data-driven and transparent urban planning in complex urban environments.</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

Harnessing hybrid intelligence and explainable AI for urban growth prediction: A Data-Driven framework for sustainable cities

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年9月20日

查看全文
  • Danish Khan,
  • Nizamuddin Khan,
  • Sajid Ullah

摘要

Urban growth modelling is crucial for sustainable development, especially in rapidly urbanizing regions where balancing infrastructure expansion with environmental sustainability poses ongoing challenges. Traditional machine learning (ML) models often fall short in capturing the complex, non-linear patterns of urban systems. To address this, we propose a hybrid intelligence framework that integrates linear, tree-based, and neural network models to enhance predictive accuracy and interpretability. Using geospatial and socio-economic datasets for Lucknow City, India, we developed three novel hybrid models: Logistic Regression-XGBoost-Feedforward Neural Network (LR-XGB-FNN), ElasticNet-LightGBM-TabNet (EN-LGBM-TN), and Generalized Linear Model-CatBoost-Wide & Deep Network (GLM-CB-WDN). These models were trained on urban growth patterns between 2011 and 2021, derived from Landsat imagery and thematic indices, and validated through change detection analysis. Among them, LR-XGB-FNN achieved the highest accuracy (94.17%) and AUC (0.9475). Spatial probability maps were generated to forecast urban expansion hotspots for 2031, offering actionable insights for planners. To improve transparency, SHAP (SHapley Additive exPlanations) was used to interpret model outputs, revealing Drive Time (DT), Elevation, and Distance from Roads (DfR) as key drivers. This study contributes to the field by (1) proposing robust hybrid ML models, (2) applying explainable AI for interpretability, (3) producing future urban growth maps, and (4) presenting a scalable framework aligned with SDG 11. The results demonstrate how hybrid intelligence can support data-driven and transparent urban planning in complex urban environments.