Efficient Data-Driven Machine Learning Models for Water Quality Prediction

Water is a valuable, necessary and unfortunately rare commodity in both developing developed countries all over the world. It undoubtedly most important natural resource on planet constitutes an essential nutrient for human health. Geo-environmental pollution can be caused by many different types of waste, such as municipal solid, industrial, agricultural (e.g., pesticides fertilisers), medical, etc., making water unsuitable use any living being. Therefore, finding efficient methods to automate checking suitability great importance. In context this research work, we leveraged supervised learning approach order design accurate possible predictive models from labelled training dataset identification suitability, either consumption or other uses. We assume set physiochemical microbiological parameters input features that help represent water’s status determine its class (namely safe nonsafe). From methodological perspective, problem treated binary classification task, machine models’ performance (such Naive Bayes–NB, Logistic Regression–LR, k Nearest Neighbours–kNN, tree-based classifiers ensemble techniques) evaluated with without application balancing (i.e., nonuse Synthetic Minority Oversampling Technique–SMOTE), comparing them terms Accuracy, Recall, Precision Area Under Curve (AUC). our demonstration, results show Stacking model after SMOTE 10-fold cross-validation outperforms others Accuracy Recall 98.1%, 100% AUC equal 99.9%. conclusion, article, framework presented support researchers’ efforts toward quality prediction using (ML).