AI Enabled Water Well Predictor

Groundwater has long served as a critical resource for agriculture, domestic use, and industry, particularly in regions lacking reliable surface water. However, the unpredictability of well yields has made sustainable extraction and planning increasingly difficult. Traditional well yield estimation methods, such as manual surveys and pump testing, are time-consuming, subjective, and lack scalability across diverse geological regions. This study introduces a machine learning-based water well yield predictor that leverages historical borewell data and site-specific features such as well depth, aquifer type, soil composition, and static water level. Using ensemble models, such as Random Forest, XGBoost, the system provides real-time predictions of both categorical yield levels and continuous flow rates. The models are trained with k-fold cross-validation and enhanced by correlation-based feature selection. Experimental results show 89% accuracy in classification and an R² of 0.93 for regression, significantly reducing the risk of low-yield drilling. The solution is deployable through a web interface and designed for continuous learning, supporting integration with GIS and climate data sources to aid long-term groundwater management.