Improving Water Quality Prediction and Monitoring Using Machine Learning Algorithms

Within the framework of growing worldwide environmental apprehensions, the significance of conserving water resources and maintaining ecological balance has been progressively evident. Consequently, the surveillance and forecasting of water quality have become crucial endeavors in attaining these goals. Nevertheless, the task of guaranteeing the efficiency and reliability of water quality forecasting has presented itself as a formidable undertaking. Therefore, this research presents an innovative approach utilizing machine learning, i.e., tree-based and ensemble learning approaches for detecting the quality of water. The dataset employed for the investigation is the publicly available Kaggle dataset. Upon analyzing the dataset, it has been determined that the dataset contains null values and has an uneven distribution of samples of different categories. In order to resolve this matter, the absent values are imputed and substituted with the mean value. In order to achieve a balanced dataset, the SMOTE approach is employed. The pre-processed data is subsequently fed into five machine learning models, namely Decision Tree, Random Forest, XGBoost, Catboost, and Adaboost. Out of all the machine learning algorithms, Catboost demonstrates superior performance by attaining a testing F1 score of 76%. The utilization of this framework enables the efficient prediction of prospective water quality and the provision of technical assistance for emergency pollution management.