Hybrid Fuzzy Regression–Artificial Neural Network for Improvement of Short-Term Water Consumption Estimation and Forecasting in Uncertain and Complex Environments: Case of a Large Metropolitan City

Abstract

This study presents a hybrid approach consisting of artificial neural network (ANN), fuzzy linear regression (FLR), and analysis of variance (ANOVA) for improvement of water consumption forecasting. Hence, this approach can be easily applied to uncertain or certain, or complex environments given its flexibility. The proposed hybrid approach is applied to forecast short-term water consumption in Tehran, Iran from April 5, 2004, to March 21, 2009. In this study, daily water consumption is viewed as the resultant of future and historical meteorological data. Implementation of the hybrid approach in a large metropolitan city such as Tehran seems to be ideal because of potential nonlinearity and uncertainty in the water consumption function of Tehran, Iran. The results of mean absolute percentage error (MAPE) indicate that selected ANN outperforms selected FLR on warm days. However, both ANN and FLR are ideal for cold days. To verify and validate the results, a sensitivity analysis is carried out by changing the train and test data sets. Finally, the comparison of the MAPE results of the hybrid approach with conventional linear regression confirms its considerable superiority for both warm and cold days.