FORECASTING SUSPENDED SEDIMENT LOAD USING REGULARIZED NEURAL NETWORK: CASE OF THE SEBAOU RIVER (ALGERIA)

Abstract

In the management of water resources in different hydro- systems it is important to evaluate and predict the sediment load in rivers. It is difficult to obtain an effective and fast estimation of sediment load by artificial neural network without avoiding over-fitting of the training data. The present paper comprises the comparison of a multi-layer perception network using the regularized neural network using the Early Stopping technique to estimate and forecast suspended sediment load in the Sebaou River, northern Algeria. The study was carried out on daily sediment discharge and water discharge data of 9 years (1978–1987).the data sets were divided into three sets; training, validation and testing. The validation set were used only for cross validation. The results of the Back Propagation based models were evaluated in terms of the coefficient of determination (R²) and the root mean square error (RMSE). The comparison results indicate that the regularizing ANN using the Early Stopping technique to avoid over-fitting stops in the first 50 epochs from 200 epochs. and the results show that the overtraining in the back propagation occurs because of the complexity of the data introduced to the network.