Extreme challenges in urban water supply are arising from the combination of growing demands, limited water resources, and climate change impacts. Considering the limited implementation and its complexity in water infrastructure management, water demand management has received increasing attention to address current water systems’ challenges. In this regard, providing feedback information on future water use, compared with other houses or buildings, will suggest the opportunity to encourage voluntary water saving to users. Numerous studies have been conducted to improve water demand prediction, recently using machine learning models with features related to water usage and climate factors. However, social, water supply infrastructure, and energy factors (e.g., building size, energy consumption, number of populations, and ageing) are also critical for urban water demand prediction. In this regard, this study tested machine learning models to predict water demand with features related to water/energy use, climate, social, and infrastructure, with a study area in Illinois in the United States. This study employed multiple models such as artificial neural network, support vector machine, random forest, and recurrent neural network. The models were trained and tested using the data for water and energy usage in the study area and the data from public data sources for climate, social, and infrastructure factors. Their prediction performance was evaluated using statistical indicators such as root mean square errors. The results showed the effects of additionally considering social, infrastructure, and energy features on water demand prediction using machine learning models. The findings and discussion will suggest insights into how to advance the existing water demand prediction models contributing to urban water demand management.
