Integrating Machine Learning Techniques to Assess and Predict Water Contamination in the Gomti River Basin

Abstract

The increasing contamination of river systems due to rapid urbanization, industrial discharge, and agricultural runoff poses a serious threat to environmental and public health. Traditional methods often fall short in capturing the complex, non-linear interactions among multiple water quality parameters. To address this gap, the present study applies advanced machine learning (ML) models to assess and predict the water quality index (WQI) in the Gomti river basin during the pre-monsoon season, a period with minimal dilution effects. A total of 100 samples were collected from five strategic sites, and 18 physicochemical and heavy metal parameters were analyzed. The study compares the performance of multiple linear regression (MLR), artificial neural networks (ANN), support vector regression (SVR), random forest regression (RFR), and decision tree regression (DTR). Results show that ANN achieved the highest predictive accuracy (92.52% of predictions within ±20% of actual WQI), followed by RFR (88%), while MLR and DTR showed limited performance. Feature importance and sensitivity analysis identified electrical conductivity (EC) dissolved oxygen (DO), total dissolved solids (TDS), and sulfate (SO₄) as the most influential predictors of WQI. This study demonstrates the potential of ML-based models for accurate water quality prediction and supports data-driven strategies for sustainable water resource management. Future work should incorporate seasonal variations, real-time sensor data, and hybrid modeling frameworks to enhance predictive reliability and support early-warning systems.