Harnessing the Power of Random Forest for Precise Short-Term Water Demand Forecasting in Italian Water Districts
Year: 2024
Type: conference-paper
Abstract: Water demand forecasting is essential for ensuring a reliable water supply in any water utility. It involves making accurate predictions for both short- and long-term water needs. Many traditional time series forecasting methods are presently used; however, recent machine learning techniques have grown in popularity for their robustness and accuracy. Random forest is an emerging machine learning algorithm which was used to forecast short-term water demand for ten district metered areas in Italy. Our predictions on test datasets using the trained model yielded correlations as high as 0.98. Important explanatory variables affecting model performance included consumption patterns represented by the seven-day water demand lag. In this paper, we present a reliable application of the random forest algorithm for short-term water demand forecasting.
Authors: Adam Kulaczkowski, Juneseok Lee
Institutions: CDM Smith (United States), Manhattan University
Language: English
Subfield: Environmental Engineering
Field: Environmental Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4402319558