Hodrıck-Prescott Filtreleme ve Yapay Sinir Ağları Yöntemiyle Uzun Dönem Su Tüketimi Tahmini: Türkiye Üzerine Uygulama

Abstract

Water is one of the most essential building blocks for living organisms. It plays an active role in vital bodily functions of organisms and is involved in reactions such as dissolution, purification, and hydrolysis in the non-living environment. In various human activities such as production, mining, cooling processes, and firefighting, water holds a critical role. Events such as global warming, increasing temperatures, population growth, and water pollution lead to a decrease in water resources. This necessitates a more efficient utilization of water resources. Water consumption forecasting becomes crucial for this purpose. Short-term forecasts aid in the efficient operation and management of existing water systems, while long-term forecasts assist in decision-making for new investments, water system planning, and expansion. This study focuses on long-term water consumption forecasting for Türkiye, aiming to calculate the country's future water needs to prevent potential water shortages. Various machine learning and expert system methods have been employed in the literature for water consumption forecasting. Among these methods, artificial neural networks, which are widely used, have been applied to this study. After determining the factors influencing water consumption, data on past water consumption quantities and influencing factors have been collected. These data has been rendered stationary by eliminating fluctuations using the Hodrick-Prescott (HP) filtering method and removing seasonality. The data obtained through HP filtering have been subjected to artificial neural networks and multiple linear regression methods to determine the model yielding the best results. Future water consumption forecasting for Türkiye has been conducted using the identified model. The forecasts indicate an increase in water consumption in the future. Some decisions that could be taken to address this increase have been anticipated.