Optimal prediction of the Nile River flow using neural networks

Abstract

Two models for forecasting the Nile River flow have been developed. A traditional linear autoregressive (AR) model and a feedforward neural networks (NNs) model are presented. A number of NNs models with variable number of neurons in the hidden layer were developed. The network with minimum training and testing normalized root mean square error was selected as the optimal network for forecasting. The performance of both the AR and NNs models were tested using a set of measurements recorded at Dongola station in Egypt. A significant improvements of the error when using NNs model was achieved.