Prediction of the Groundwater Levels Based on Random Forest Regression Algorithm

Random forest regression is a widely used machine learning algorithm. In this study, random forest regression is employed to predict groundwater levels. Five influencing factors are considered: river flow, temperature, saturation deficit, precipitation, and evaporation capacity. Each factor is regarded as a feature. Twenty-four sets of data were collected: eighteen sets for training and six for testing. The groundwater levels predicted using the proposed approach agree well with the corresponding true values. The relative errors are less than 5%, which satisfies the requirements of engineering precision. A feature importance analysis is also conducted. The precipitation has the most significant influence on the groundwater levels, with a degree of more than 50%, and the saturation deficit has the least influence, less than 10%. The results of this study not only provide an effective approach for groundwater level prediction but also demonstrate the encouraging prospects of machine learning algorithms in hydrological engineering applications.