MULTIVARIATE ASSESSMENT AND MACHINE LEARNING-BASED PREDICTION OF WATER POLLUTION USING PHYSICOCHEMICAL PARAMETERS FOR SUSTAINABLE ENVIRONMENTAL MONITORING

Abstract

This study presents an integrated approach for the assessment and prediction of water pollution using physicochemical parameters through multivariate statistical techniques and machine learning models. Water quality degradation caused by rapid urbanization, industrial discharge, and agricultural activities necessitates efficient monitoring and predictive frameworks. In this study, key physicochemical parameters including pH, temperature, dissolved oxygen (DO), biological oxygen demand (BOD), chemical oxygen demand (COD), total dissolved solids (TDS), and turbidity were analyzed to evaluate water quality. Multivariate statistical methods such as correlation analysis, Principal Component Analysis (PCA), and cluster analysis were employed to identify relationships among parameters, determine dominant pollution sources, and classify sampling locations based on similarity. The results indicated strong correlations among pollution indicators, with BOD, COD, and TDS emerging as primary contributors to water contamination. Machine learning models including Linear Regression, Decision Tree, Support Vector Machine, and Random Forest were developed to predict water quality. Among these, the Random Forest model demonstrated superior performance with higher accuracy and lower prediction errors. Feature importance analysis further confirmed the significant influence of organic and chemical pollutants on water quality. The integration of statistical and machine learning approaches provides a robust and reliable framework for water quality assessment and prediction. This study highlights the potential of data-driven methodologies for sustainable environmental monitoring and supports informed decision-making for effective water resource management.