Scholarlink

<p>With the continuous increase in urban growth, rising land demand can encroach on valuable areas and develop in hazardous zones, such as flood-prone regions, leading to environmental degradation, ecological imbalances, and social challenges. This study addresses these issues by integrating machine learning and deep learning with spatial decision-making to predict urban growth in Bou Saâda, Algeria. Two independent predictive models, XGBoost and Deep Neural Networks (DNN), were developed and trained on historical land-use changes between 1990 and 2006, and validated using data from 2006 to 2022. The results indicate that DNN achieved a prediction accuracy of 95%, slightly outperforming XGBoost at 93%. Contrary to previous studies that often report superior performance of XGBoost over DNN, our findings suggest that the spatial nature and complexity of the dataset may favor the nonlinear modeling capacity of DNN. To assess the spatial suitability of future urban growth, the Analytic Hierarchy Process (AHP) was applied, integrating geographic and infrastructural factors to identify optimal development zones. Comparing the predicted urban growth with suitability results allowed evaluation of whether future growth aligns with suitable areas or poses environmental and infrastructural risks. The findings highlight potential threats from unplanned urban growth, particularly to natural vegetation, oases, and flood-prone valleys. Mitigation strategies include GIS- and AI-based monitoring, community engagement, and data-driven planning frameworks. Despite limitations related to feature selection, dataset quality, and focus on a single city, the integrated modeling and suitability assessment provide an effective decision-support tool for sustainable urban growth planning and future research directions.</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

Modeling and evaluation of urban growth: a comparative approach between XGBoost and deep neural networks coupled with multi-criteria analysis: case of BouSaâda, Algeria

  • 期刊论文

  • 发表日期：

    2025年11月15日

查看全文
  • Sadiq Tahmi,
  • Xiaoping Rui,
  • Imen Guechi

摘要

With the continuous increase in urban growth, rising land demand can encroach on valuable areas and develop in hazardous zones, such as flood-prone regions, leading to environmental degradation, ecological imbalances, and social challenges. This study addresses these issues by integrating machine learning and deep learning with spatial decision-making to predict urban growth in Bou Saâda, Algeria. Two independent predictive models, XGBoost and Deep Neural Networks (DNN), were developed and trained on historical land-use changes between 1990 and 2006, and validated using data from 2006 to 2022. The results indicate that DNN achieved a prediction accuracy of 95%, slightly outperforming XGBoost at 93%. Contrary to previous studies that often report superior performance of XGBoost over DNN, our findings suggest that the spatial nature and complexity of the dataset may favor the nonlinear modeling capacity of DNN. To assess the spatial suitability of future urban growth, the Analytic Hierarchy Process (AHP) was applied, integrating geographic and infrastructural factors to identify optimal development zones. Comparing the predicted urban growth with suitability results allowed evaluation of whether future growth aligns with suitable areas or poses environmental and infrastructural risks. The findings highlight potential threats from unplanned urban growth, particularly to natural vegetation, oases, and flood-prone valleys. Mitigation strategies include GIS- and AI-based monitoring, community engagement, and data-driven planning frameworks. Despite limitations related to feature selection, dataset quality, and focus on a single city, the integrated modeling and suitability assessment provide an effective decision-support tool for sustainable urban growth planning and future research directions.