Forecasting of monthly river flow with autoregressive modeling and data-driven techniques

Terzi O., Ergin G.

NEURAL COMPUTING & APPLICATIONS, vol.25, no.1, pp.179-188, 2014 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 25 Issue: 1
  • Publication Date: 2014
  • Doi Number: 10.1007/s00521-013-1469-9
  • Journal Name: NEURAL COMPUTING & APPLICATIONS
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.179-188
  • Süleyman Demirel University Affiliated: Yes

Abstract

This study was conducted by using autoregressive (AR) modeling and data-driven techniques which include gene expression programming (GEP), radial basis function network and feed-forward neural networks, and adaptive neural-based fuzzy inference system (ANFIS) techniques to forecast monthly mean flow for KA +/- zA +/- lA +/- rmak River in Turkey. The lagged monthly river flow measurements from 1955 to 1995 were taken into consideration for development of the models. The correlation coefficient and root-mean-square error performance criteria were used for evaluating the accuracy of the developed models. When the results of developed models were compared with flow measurements using these criteria, it was shown that the AR(2) model gave the best performance among all developed models and the GEP and ANFIS models had good performance in data-driven techniques.