A Comparison of MLR and Three Different Artificial Neural Networks Techniques for Daily Mean Flow Prediction

Abstract

Prediction for hydrologic events has always been an important issue for optimizing and planning the whole system. In this study, a conventional multivariate linear regression method and three different ANN (Artificial Neural Networks) techniques (Feed Forward Back Propagation, Generalized Regression and Radial Basis Function) were used to predict and model daily mean flow of Anamur River. Because of the climate change, last 5234 days' data between the years 1989 and 2003 were processed and those three techniques of ANN and the conventional method MLR (multivariate linear regression) were compared to each other. The best results were obtained with FFBP algorithm.