Evaluation of Moving Average Pre-processing Approach to Improve the Efficiency of Support Vector Regression Model for Inflow Prediction

Accurate hydrologicalforecasting is a main tool for the water resources planning. In this paper, the inflow rates to Bakhtiari and Rudbar Dams in Lorestan province – IRAN, were forecasted using support vector regression (SVR), Multiple Linear Regression (MLR) and Autoregressive Moving Average (ARMA) models. In order to pre-process the input data for the above mentioned models, the moving average approach was used. Furthermore, Nash-Sutcliffe Efficiency (NSE), Root Mean Square Error (RMSE), correlation coefficient (R) and Taylor diagram were used to evaluate the efficiency of the models. The results showed that the moving average pre-processing approach improved the performance of the above mentioned models dramatically. For instance, the values of Nash-Sutcliff correspond to SVR hybrid model in forecasting inflow rate to Bakhtiari and Rudbar-Lorestan dams with moving average pre-processing were improved by 13.4% and 6.6%, respectively, as compared to those in the SVR model without pre-processing.