Machine Learning Framework for Urban Water Demand Forecasting: A Wavelet-Enhanced Support Vector Regression Ensemble

This study proposes a novel hybrid model for short-term urban water demand (UWD) forecasting in the Punjab region of India, coupling support vector regression (SVR) with wavelet transform (WT) and boosting ensemble techniques. The WTs are applied as a preprocessing step to decompose the water demand time series into multiple frequency components capturing both long-term trends and short-term variations. By integrating wavelet analysis with boosting ensemble, the accuracy and robustness of the SVR models were significantly improved. The model’s performance was assessed for lead times of 1, 3, and 5 days ahead using the coefficient of determination ( R 2 ), root mean square error, and mean absolute error. Results showed that boosting consistently enhanced the correlation between observed and predicted UWD values. Among the models evaluated, the wavelet-boosting SVR outperformed others, providing the most accurate forecasts. This highlights the potential of wavelet-based boosting techniques for improving UWD prediction and aiding in efficient water resource management.