Does restraining end effect matter in EMD-based modeling framework for time series prediction? Some experimental evidences
Demand forecasting after natural disasters is especially important in emergency management. However, since the time series of commodities demand after natural disasters usually has a great deal of nonlinearity and irregularity, it has poor prediction performance of applying the traditional statistical and econometric models such as linear regression and autoregressive moving average (ARMA) to this kind of data. This paper tries to apply a hybrid forecasting method which is an integration of empirical mode decomposition (EMD) and autoregressive integrated moving average (ARIMA). The EMD-ARIMA forecasting methodology is then applied to the prediction of agricultural products demand after the 2008 Chinese winter storms. Forecasting results indicate that EMD can improve the prediction accuracy of classical ARIMA forecasting method for demand of commodities after natural disasters. 2009 Elsevier Ltd. All rights reserved.