Enhancing prediction accuracy and data handling for environmental applications in innovative modeling of underground water level fluctuations based on tree ensembles technique

Abstract

This study developed a model to evaluate and predict underground water level fluctuations based on various factors that affect water reserves. The process of calculating input data features was performed to improve forecast quality. The paper emphasizes the automatic handling of missing and noisy data before incorporating them into the training dataset. Subsequently, the Tree Ensembles learning method was applied to construct the underground water level prediction model. The results indicate that the model can accurately predict the trend of changes in water level in water storage areas such as aquifers and lakes. In particular, this method demonstrated flexibility in handling various input variables, including erroneous, missing, and noisy data, without requiring overly complex preprocessing. This opens up the potential for applying underground water level prediction models in real-world scenarios, where data is often highly diverse and complex. In conclusion, this study not only provides an effective method to predict fluctuations in the level of underground water at storage points, but also suggests significant potential for the development of evaluation and prediction models in the environmental field in the future.