Double trends time series forecasting using a combined ARIMA and GMDH model

The time series of monthly cigarette sales have double trends which include long-term upward trend and seasonal fluctuations trend. For this complex system forecasting, single linear or nonlinear forecasting model can't deeply capture characteristics of the data so the results are imprecise. In this paper, a combined methodology that combines both ARIMA and GMDH models is proposed to take advantage of the unique strength of ARIMA and GMDH models in linear and nonlinear modeling. These two models are combined based on info entropy method. Experimental results with real data sets indicate that the proposed combined model can be an effective way to improve forecasting accuracy achieved by either of the models used separately.