Abstract

The water resources are gradually decreasing in the world due to reasons such as decrease in precipitation result of climate changes, epidemics, unconscious water consumption, improper irrigation problems, industrialization and population increases. Economically, considering the widespread use of water in agriculture and industry, it may cause crises between countries in future. For this reason, it is very important for countries to predict the amount of water consumption and to create appropriate water policies in the light of the predictions. The prediction of water consumption is a critical issue due to ever-increasing population and the risk of insufficient water resources of Istanbul. Therefore, in this study, in order to investigate the best prediction model of water consumption, ARIMA from classical time series methods, machine learning methods and hybrid models, which are alternative methods that have emerged recently and have shown quite good performance in predicting both linear and nonlinear structures in time series, are applied. Firstly, stationarity analysis are made by applying augmented Dickey Fuller (1979,1981), Phillips and Perron (1988) and Kwiatkowski, Phillips, Schmidt and Shin (1992) unit root tests to the yearly drinking water consumption series in 1991-2022 period. To determine the best estimation model for water consumption in Istanbul, we use the hybrid model of Zhang (2003), which allows us to use ARIMA from time series, MLP (Multi-Layer Perceptron) from artificial neural networks, and artificial neural networks for linear data and non-linear time series data. The water consumption of İstanbul is predicted by using the ARIMA, MLP (Multilayer Perceptron) and Zhang (2003)'s hybrid model, and the results are compared based on error-based criteria. As a result, Zhang (2003)'s hybrid model is determined as the best estimation model. In addition, based on the prediction results obtained from Zhang (2003)'s hybrid model, it is concluded that Istanbul's water consumption will continue to increase in coming years.