A Software Application – Ground Water Level Predictor

Abstract: Groundwater is a major source of water for drinking, farming and industry, so knowing how its level will change is very important. Many studies have shown that machine learning and deep learning can predict groundwater levels well, but most of them focus on building and comparing models. Few of them offer a complete and easy-to-use software application. This paper presents a Ground Water Level Predictor, a software application that brings data input, data cleaning, prediction and result display together in one system. The application checks the uploaded data, fills missing values, creates useful features such as previous groundwater levels, rainfall and month, and trains three models: Linear Regression, Random Forest and Gradient Boosting. It then selects the model with the lowest error and uses it to predict the groundwater level. The results are shown as charts in a web-based interface, so no programming knowledge is needed. The system was demonstrated on a dummy monthly dataset of 120 records, where Linear Regression performed best with an RMSE of 0.205 m and an R² of 0.980. The application shows that groundwater prediction can be made simple and practical for users, and it can be tested next with real well data.

Keywords: Groundwater level prediction; software application; machine learning; Linear Regression; Random Forest; Gradient Boosting; data preprocessing; user interface.