Water Foot Print Calculator using Machine Learning

Water scarcity is becoming an increasingly critical challenge, particularly in developing countries like India, where per capita water availability is rapidly declining. Accurate estimation of domestic water footprints is essential for sustainable water resource management, yet real-time household water consumption data remains scarce. This study proposes a machine learning-based framework, utilizing the XG Boost algorithm, to predict daily household water consumption from structured activity-based features such as household size, activity type, activity duration, and frequency. Due to the lack of publicly available datasets, asynthetic dataset was generated and validated against reports from the National Sample Survey Office (NSSO) and the Central Water Commission (CWC). The model achieved outstanding performance with an R² score of 0.94, RMSE of 4.5 liters/day, and MAE of 3.2 liters/day, outperforming Random Forest, Artificial Neural Networks, and traditional regression models. Feature importance analysis revealed that toilet flushing, showering, and gardening are the highest contributors to household water consumption. The study concludes that machine learning techniques, even when applied to synthetic data, can provide accurate, interpretable, and scalable solutions for domestic water footprint estimation. Future work will focus on integrating IoT- based real- time monitoring, demographic-based modeling, and deep learning approaches to further enhance prediction accuracy and support sustainable water management initiatives.