Scholarlink

<p>This study presents a comprehensive AI-driven framework for assessing and simulating the impact of Piped Water Supply (PWS) on poverty risk across rural habitations in Rajasthan, India-advancing Sustainable Development Goal 1 (No Poverty). A four-phase methodology integrates data preprocessing, supervised machine learning, causal inference and real-time policy simulation. Using habitation-level data from multiple districts, key features such as SC/ST population ratios and water connectivity were engineered. Among several classification models, XGBoost outperformed others with an Area Under the Curve (AUC) of <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\:0.9916\)</EquationSource> </InlineEquation> and an F1-score of <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\:0.7659\)</EquationSource> </InlineEquation>. SHAP analysis confirmed (SC + ST) %, ST population and PWS connections per household as critical predictors. Causal inference revealed that PWS access causally reduces poverty risk by an average of <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(\:11.3\)</EquationSource> </InlineEquation> percentage points, confirmed via placebo refutation. A Streamlit dashboard enables interactive simulation of counterfactual scenarios allowing users to estimate poverty risk reductions under universal PWS coverage. The tool translates predictive and causal outputs into actionable insights for infrastructure planning, offering scalable, real-time support for targeted policy interventions. This integrated approach bridges the gap between AI modeling and field-level decision-making offering a replicable model for data-informed rural development.</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

AI-driven Poverty Risk Mapping and Causal Assessment of Piped Water Supply in Rural Region: A Policy Simulation Framework Aligned with SDG 1

  • 期刊论文

  • 发表日期：

    2026年3月28日

查看全文
  • Kezia Saini,
  • Priyam Nath Bhowmik

摘要

This study presents a comprehensive AI-driven framework for assessing and simulating the impact of Piped Water Supply (PWS) on poverty risk across rural habitations in Rajasthan, India-advancing Sustainable Development Goal 1 (No Poverty). A four-phase methodology integrates data preprocessing, supervised machine learning, causal inference and real-time policy simulation. Using habitation-level data from multiple districts, key features such as SC/ST population ratios and water connectivity were engineered. Among several classification models, XGBoost outperformed others with an Area Under the Curve (AUC) of \(\:0.9916\) and an F1-score of \(\:0.7659\) . SHAP analysis confirmed (SC + ST) %, ST population and PWS connections per household as critical predictors. Causal inference revealed that PWS access causally reduces poverty risk by an average of \(\:11.3\) percentage points, confirmed via placebo refutation. A Streamlit dashboard enables interactive simulation of counterfactual scenarios allowing users to estimate poverty risk reductions under universal PWS coverage. The tool translates predictive and causal outputs into actionable insights for infrastructure planning, offering scalable, real-time support for targeted policy interventions. This integrated approach bridges the gap between AI modeling and field-level decision-making offering a replicable model for data-informed rural development.