Ecology, Environment and Conservation Paper

A COMPARATIVE STUDY OF ARTIFICIAL NEURAL NETWORKS AND SUPPORT VECTOR REGRESSION MODELS FOR STREAMFLOW PREDICTION

Behrouz Yaghoubi and Ahmad Rajabi

Abstract

In this study, the monthly rainfall-runoff of Gavehrood in Kermanshah province was investigated. Support vector regression (SVR) was used to predict the stream flow. The results were compared with those obtained from multilayer perceptron (MLP) artificial neural networks (ANNs). For this purpose, the 49-year (1960-2009) monthly precipitation, temperature and runoff data at the station entrance to the storage dam at the outlet of the basin were analyzed. The results were evaluated by RSR, NSC and CC indicators. Comparison of the results showed that SVR predicts non-linear behavior of flow data with a higher accuracy than ANN model. Accordingly, this model can be used as a promising and reliable forecasting tool for rainfall-runoff modeling.