Scholarlink

India, a country with abundant water resources, holds only 4% of the world’s total water resources. However, a significant portion of its rivers, lakes, and surface water bodies are heavily polluted due to untreated sewage, industrial waste, and solid waste. In light of this, a research effort is proposed to assess the Water Quality Index (WQI) of the Narmada River, a prominent river in India, and determine its portability. The research utilizes a real-time dataset of water quality parameters, including pH, total coliforms, biochemical oxygen demand, electrical conductivity, nitrate, fecal coliforms, fecal streptococci, and temperature, collected from several rivers in India in recent years. In the proposed work, the water quality data are utilized by applying various preprocessing techniques to predict the Water Quality Index (WQI) through regression methods like Multi-Linear Regression, ElasticNet Regression, and Random Forest (RF) Regression. In addition to predicting the WQI, the given samples are also classified into suitable and non-suitable types of water based on WQI. Several machine learning (ML) approaches, such as RF, XGBoost, Logistic Regression (LR), Decision Tree, and K-Nearest Neighbour, are used for this classification task. The dataset is balanced using the Synthetic Minority Oversampling Technique (SMOTE), which is essential since it is unbalanced. The classification results reveal that LR provides the highest accuracy of 100%, while for regression, RF Regressor is the most efficient algorithm, with an R-squared value of 0.91 and an RMSE value of 0.13. The study aims to establish a baseline for future research on water quality in Indian rivers and also provides decision-makers with crucial information on appropriate sampling and analytical techniques to manage the impact of pollution on river surface water quality.

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-Based Water Quality Prediction Using Machine Learning and Synthetic Minority Oversampling Technique

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sarthak Kapaliya,
  • Kaxit Pandya,
  • Luv Patel,
  • Dhruvil Patel,
  • Rajeev Kumar Gupta

摘要

India, a country with abundant water resources, holds only 4% of the world’s total water resources. However, a significant portion of its rivers, lakes, and surface water bodies are heavily polluted due to untreated sewage, industrial waste, and solid waste. In light of this, a research effort is proposed to assess the Water Quality Index (WQI) of the Narmada River, a prominent river in India, and determine its portability. The research utilizes a real-time dataset of water quality parameters, including pH, total coliforms, biochemical oxygen demand, electrical conductivity, nitrate, fecal coliforms, fecal streptococci, and temperature, collected from several rivers in India in recent years. In the proposed work, the water quality data are utilized by applying various preprocessing techniques to predict the Water Quality Index (WQI) through regression methods like Multi-Linear Regression, ElasticNet Regression, and Random Forest (RF) Regression. In addition to predicting the WQI, the given samples are also classified into suitable and non-suitable types of water based on WQI. Several machine learning (ML) approaches, such as RF, XGBoost, Logistic Regression (LR), Decision Tree, and K-Nearest Neighbour, are used for this classification task. The dataset is balanced using the Synthetic Minority Oversampling Technique (SMOTE), which is essential since it is unbalanced. The classification results reveal that LR provides the highest accuracy of 100%, while for regression, RF Regressor is the most efficient algorithm, with an R-squared value of 0.91 and an RMSE value of 0.13. The study aims to establish a baseline for future research on water quality in Indian rivers and also provides decision-makers with crucial information on appropriate sampling and analytical techniques to manage the impact of pollution on river surface water quality.