Machine learning-based forecasting of potability of drinking water through adaptive boosting model

Abstract

Water is an indispensable requirement for life for health and many other purposes, but not all water is safe for consumption. Thus, various metrics, such as biological, chemical, and physical, could be used to determine the quality of potable water for use. This study presents a machine learning-based model using the adaptive boosting technique with the ability to categorize and evaluate the quality rate of drinking water. The dataset for the study was adopted from Kaggle. Consequently, an experimental analysis of the different machine learning techniques (ensemble) was carried out to create a generic water quality classifier. The results show that the forecast accuracy of the logistic regression model (88.6%), Chi-square Automatic Interaction Detector (93.1%), XGBoost tree (94.3%), as well as multi-layered perceptron (95.3%) improved by the presented ensemble model (96.4%). The study demonstrates that the use of ensemble model presents more precision in predicting water quality compared to other related algorithms. The use of the model presented in this study could go a long way to enhance the regulation of water quality and safety and address the gaps in conventional prediction approach.