1. Title, Journal and Authors
Title : Water demand forecasting: review of soft computing methods
Journal : Environ Monit Assess
Authors : Iman Ghalehkhondabi a*, Ehsan Ardjmand b, William A. Young II c, Gary R. Weckman a
a Department of Industrial and Systems Engineering, Russ College of Engineering and Technology, Ohio University, Athens, OH 45701, USA
b Department ofManagement, College of Business, Frostburg State University, Frostburg, MD 21532, USA
c Management Information Systems Department, College of Business, Ohio University, Athens, OH 45701
Research on modeling water demand forecasting through soft computing technology has been ongoing. These methods include artificial neural networks (ANNs), fuzzy and neuro-fuzzy models, support vector machines, metaheuristics, and system dynamics. This paper is described by comparing the pros and cons of each type of model. There are many types of ANN models and are sometimes sensitive to climate. In particular, the accuracy of water demand prediction is improved when using the wavelet-bootstrap ANN model. The researchers also concluded that they outperform conventional ARIMAs. System dynamics is a more mathematical model, but its application is more limited than other soft computing methods. Recently, the hybrid model, rather than the classical model method, can improve the prediction accuracy.
3. Application to research
- Recurrent neural networks used in other sectors, such as energy and finance, can help predict water demand. The ensemble method is widely used in other fields, and thus can be sufficiently benefited in the water demand forecasting field. Also, if the validity test is performed before running the model, higher accuracy results can be obtained.
- When forecasting water demand, socioeconomic variables are not commonly used for long-term forecasting.