Akademik Veri Yönetim Sistemi
Earth Science Informatics, cilt.18, sa.2, 2025 (SCI-Expanded, Scopus)
Water, one of the most important sources of life, is very important for living life. In this case, improving the quality of water resources is indispensable for life. Electrical Conductivity (EC) is one of the criteria affecting water quality. An increase in conductivity indicates a higher amount of non-water substances and suggests that the water is losing its essential properties. Therefore, accurately predicting EC greatly contributes to the management strategies of water resources. In this paper, we present GA-ML, a novel approach that utilizes a Genetic Algorithm (GA) for end-to-end hyperparameter tuning across five different machine learning methods to improve the prediction of water’s EC. The GA was employed to select the optimal machine learning algorithm in the recursive feature elimination (RFE) for feature selection. Subsequently, the best normalization method was applied to the selected features, followed by comprehensive hyperparameter optimization. The results demonstrated that the Extra Trees based GA-ML (GA-ET) model achieved the lowest root mean squared error (RMSE) of 11.8508, while XGBoost based GA-ML (GA-XGBoost) provided the best performance in terms of mean absolute error (MAE) with a value of 9.2238. To our knowledge, this is the first study to use GA for both parameter tuning and optimizing preprocessing and feature engineering. Future research will explore applying this multi-step metaheuristic approach to various prediction problems.