Scholarlink

<p>Monitoring and managing water quality (WQ) is a major concern and is essential for building sustainable cities, especially at sacred water bodies which hosts largest religious water dip gatherings. Current research leverages Ensemble Learning (EL) techniques to predict trends in WQ in 7 sites located around Triveni Sangam, Prayagraj, during The Maha Kumbh Mela, utilising a comprehensive dataset that includes WQ parameters like pH, dissolved oxygen (DO), turbidity, biochemical oxygen demand (BOD), chemical oxygen demand (DO), and fecal coliform. EL models such as XGBoost, Random Forest, Gradient Boost and AdaBoost techniques were employed to predict the WQ parameters. Among the models, XGBoost consistently achieved the highest predictive accuracy, with the lowest MAE and RMSE, for instance, 0.05 and 0.07 for pH, and 517.60 and 702.43 for fecal coliform. It was also effective for highly variable indicators like BOD, COD, turbidity. The interpretations of the models are done through the use of XAI techniques such as permutation feature importance (PFI) and partial dependence plots (PDP). PFI identified that COD and fecal coliform are the factors influencing BOD, whereas DO was found to be dependent on flow and turbulence. The insights from the PDP indicated that high values of BOD and COD contribute to pollution, whereas low values of DO and high BOD suggest significant degradation of the WQ. This demonstrates the application of XAI techniques not only helps in interpreting ML model predictions of water pollutant dynamics but also supports well-informed decision-making concerning water resource management in future mass water dip events, which will help create sustainable cities of the future that are linked to sacred water bodies.</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

Explainable AI for Water Quality Intelligence in Sacred Waters: Insights from the World’s Largest Water Dip Gathering

  • 期刊论文

  • 发表日期：

    2026年7月11日

查看全文
  • Mantena Sireesha,
  • Abdul Gaffar Sheik,
  • Ameer Khan Patan,
  • Upaka Rathnayake

摘要

Monitoring and managing water quality (WQ) is a major concern and is essential for building sustainable cities, especially at sacred water bodies which hosts largest religious water dip gatherings. Current research leverages Ensemble Learning (EL) techniques to predict trends in WQ in 7 sites located around Triveni Sangam, Prayagraj, during The Maha Kumbh Mela, utilising a comprehensive dataset that includes WQ parameters like pH, dissolved oxygen (DO), turbidity, biochemical oxygen demand (BOD), chemical oxygen demand (DO), and fecal coliform. EL models such as XGBoost, Random Forest, Gradient Boost and AdaBoost techniques were employed to predict the WQ parameters. Among the models, XGBoost consistently achieved the highest predictive accuracy, with the lowest MAE and RMSE, for instance, 0.05 and 0.07 for pH, and 517.60 and 702.43 for fecal coliform. It was also effective for highly variable indicators like BOD, COD, turbidity. The interpretations of the models are done through the use of XAI techniques such as permutation feature importance (PFI) and partial dependence plots (PDP). PFI identified that COD and fecal coliform are the factors influencing BOD, whereas DO was found to be dependent on flow and turbulence. The insights from the PDP indicated that high values of BOD and COD contribute to pollution, whereas low values of DO and high BOD suggest significant degradation of the WQ. This demonstrates the application of XAI techniques not only helps in interpreting ML model predictions of water pollutant dynamics but also supports well-informed decision-making concerning water resource management in future mass water dip events, which will help create sustainable cities of the future that are linked to sacred water bodies.