Likewise, Choi et al. (2011) developed a hybrid forecasting scheme, which combines SARIMA model and wavelet transform (SW) and was shown to perform better than classical SARIMA model, classical seasonal decompositionþlinear extrapolation seasonal adjustment method (CSDþLESA) and evolutionary neural networks (ENN) based on forecast accuracy and computational time. Nie et al. (2011) introduced a hybrid model which integrates ARIMA with support vector machines (SVM) to forecast short-term load forecasting for energy management. Aye et al. (2015) compared 26 (23 individual and 3 combination) forecasting models using monthly aggregated retail sales in South Africa. They found that the combination forecasts offer ways of incorporating and discarding information from a larger number of forecasting models. Specifically, the discounted combination forecast model (DISC) outperforms all the individual models, and the other two combination forecast (simple mean and principal component) models. Pektaş and Kerem Cigizoglu (2013) proposed a hybrid ANNseasonal decomposition model to forecast monthly runoff coefficients to overcome the drawbacks of ARIMA, ARIMAX and ANN. Kriechbaumer et al. (2014) highlighted the use of combined wavelet-ARIMA approach for forecasting monthly prices of aluminum, copper, lead and zinc to improve the forecasting performance of individual models.