Ensemble learning-based approach for water quality prediction

Water quality prediction is crucial for environmental protection and water resource management, particularly in accurately predicting pH values of water bodies, which can help identify potential water quality issues and take appropriate measures. This study utilizes historical water quality data provided by the United States Geological Survey to predict the pH values of multiple water stations in Georgia for the following day. Various feature engineering methods, such as Principal Component Analysis and Lasso Regression, along with machine learning models, were applied. We combined the strengths of different models using the Stacking ensemble learning technique, which significantly improved predictive performance. The experimental results show that the combination of feature engineering— specifically Polynomial Feature Generation—and the Stacking ensemble model effectively enhanced prediction accuracy. This research provides a new approach to improving water quality monitoring methods and contributes to more precise water resource management and environmental protection.