Forecasting urban water consumption using bayesian networks and gene expression programming

Forecasting Urban Water Consumption (UWC) has a significant impress in efficient urban water management in rapidly growing cities in arid regions. Enhancing forecasting accuracy of UWC using novel models is a crucial requirement in order to the management of smart cities. In this study, Bayesian Networks (BN) is developed as a probabilistic model and compared to Gene Expression Programming (GEP) model as an evolutionary algorithm for forecasting UWC. The amount of current water consumption predicts future water consumption. The scenario with sunshine hours was added to the parameter set as the best scenario in both BN and GEP models based on comparison of Root Mean Square Error (0.11, 0.16), Mean Absolute Relative Error (0.02, 0.05), Max Root Error (0.26, 0.26), and Coefficient of determination (0.8, 0.7), respectively. The outcomes indicate that the BN model provided a more desirable efficiency compared to the GEP model. Furthermore, it can be concluded that the sunshine hour has a considerable influence on UWC, and the ability of the BN model is greatly enhanced by adding this predictor to forecast UWC in a city in an arid region with rapid population growth.