Water Quality Classification Using Machine Learning Techniques

Abstract

There is no life without water. All humans, plants, and animals need water to live. It is important to know if drinking water, a resource of human life, will be enough for everyone now and in the future. Access to clean water and hygiene is an important human right and part of the health safety policy. At the national, state, and local levels, clean water is a critical problem for health and development. This work's primary goal is to use various modeling techniques based on machine learning, deep learning, and ensemble learning to measure water quality using hyperparameter tuning of each algorithm. We have used SVM, RF, XGBoost, DT, and LGBM model stacking and voting ensemble for efficient and fast prediction. PH, chloramines, hardness, solids, sulfate, organic carbon, conductivity, trihalomethane, turbidity, and potability were the parameter used as a feature vector. A different machine, deep, and ensemble learning algorithm was used to evaluate water prediction, and the effects are compared on the accuracy, ROC AUC values, precision, recall, F1-score, MCC, and kappa score. In addition, the Freidman Ranking is also used to evaluate the model's efficiency. According to related studies, ensemble learning-based models are the most effective.