Machine Learning-Based Water Pollution Prediction for Sustainable Environmental Monitoring

Abstract

Pollution of water and the decreasing supply of clean drinking water have now become one of greatest environmental and human health issues in the world. The conventional methods of water quality assessment are usually costly, time-consuming and ineffective in continuous environmental monitoring. Here, the current paper examines application of machine learning techniques in predicting water potability using physicochemical water quality indicators. The authors used a publicly available dataset of 3276 water samples with nine environmental factors, including conductivity, organic carbon, trihalomethanes, pH, hardness, solids, sulphate, chloramines, and turbidity. Several monitored ML algorithms, which included; LR, Decision Tree, Random Forest, Support Vector Machine and K-Nearest Neighbor, were comparatively tested in terms of predictive performance. Mean imputation, feature scaling, and exploratory data analysis methods were applied to the data to prepare it before the development of the model. The results showed that SVM had highest accuracy, whereas the Decision Tree model had the most balanced classification results in terms of the F1-score. The most influential predictors of water potability were found to be pH, sulfate, solids, and hardness through feature importance analysis. The findings emphasize the suitability of ML methods in assisting intelligent environmental surveillance, sustainable water resources, and systems of early contamination. The research has a contribution to the increasing role of artificial intelligence to the environment and offers meaningful implications to policy-makers, smart cities, and health agencies.