Optimising Water Quality Assessment: A Comparative Study of Machine Learning Algorithms for Imbalanced Potability Datasets
This research demonstrates the power of machine learning in predicting water potability, showcasing improvements in prediction accuracy through hyperparameter tuning. Models like Random Forest and Gradient Boosting Machines outperform others, providing reliable insights into water safety. Hyperparameter tuning enhances model performance, optimising accuracy and generalisation. The study highlights key water quality parameters, such as pH and turbidity, for predicting potability. It also compares different algorithms to determine the best fit for water quality prediction tasks. This approach could revolutionise water monitoring, enabling real-time assessments and early contamination warnings. Machine learning models can scale globally, adapting to various water sources and conditions. This research emphasises the importance of technology in improving water management practices. Predictive models can guide policymakers in addressing water contamination risks and formulating regulations. Ultimately, machine learning offers a proactive, data-driven solution to global water quality challenges. Major Findings: Hyperparameter tuning and dataset balancing boost ML accuracy for water potability prediction. Random Forest performed best (89%), ahead of Bagging (87%) and XGBoost (83%). Optimized parameters and class imbalance handling improve reliability, suggesting deep learning potential.