Water Potability Prediction Using Machine Learning

Abstract

Abstract: For human survival, water is an essential and indispensable resource, and preserving its purity is paramount to people's health. Contaminated drinking water can lead to serious health problems, such as cholera, diarrhea, and other waterborne illnesses. Thus, maintaining clean and safe water becomes essential to advancing public health. Recent research indicates that water-related ailments claim the lives of a noteworthy 3,575,000 individuals annually. Thus, a reliable indicator of water potability could significantly lower the prevalence of these illnesses. Machine learning algorithms have emerged as highly effective instruments for precisely and promptly monitoring water resources by accurately forecasting the quality of the water. The Drinking Water dataset on Kaggle is the source of the water samples used in this study, and various algorithms are used to estimate water potability based on these properties. Nine different metrics make up this dataset: pH, hardness, solids, trihalomethanes, sulphates, chloramines, organic carbon, conductivity, and turbidity. We seek to ascertain the potability of drinking water by utilizing a variety of algorithms, including Random Forest, SVM, Decision Tree, and KNN. Among other notable results, the Random Forest algorithm outperforms conventional machine learning models, producing an astounding accuracy of 99.5%. It also performs well, producing an accuracy of 74%. As a result, this study has great potential to supply researchers, water management professionals, and policymakers with accurate data on water quality, increasing the efficacy of water potability monitoring