Enhancing Water Level Prediction through a Hybrid Feature Selection Approach

Accurate prediction of water levels (WL) is essential for various applications, from flood management to environmental monitoring. In this study, an enhanced approach to feature selection tailored for water level prediction models is presented. Our method integrates Mutual Information and Recursive Feature Elimination with Cross-Validation (RFECV), augmented by the Non-Dominated Sorting Genetic Algorithm II (NSGA-II), to systematically evaluate and refine subsets of features. Mutual Information facilitates the identification of relevant feature dependencies, while RFECV iteratively eliminates less informative features to optimize predictive accuracy. The inclusion of NSGA-II further enhances the selection process by considering multiple conflicting objectives simultaneously, such as maximizing R2 score and minimizing the number of selected features, RMSE, and MAE. Through extensive experimentation and validation on real-world datasets, we demonstrate the effectiveness of our hybrid feature selection approach in capturing intricate relationships within the data, leading to significantly improved predictive performance in water level prediction models.