Forecasting the volume of postal services using Savitzky-Golay filter modification

In this paper, an approach for investigating the time series analysis with a goal to forecast the volume of express mail services (EMS) in international traffic in Republic of Serbia is presented. We have developed new algorithm, implemented in MATLAB environment, that is based on the non-weighted symmetric Savitzky-Golay (SG) filtering and compared to simple moving average within the seasonalized autoregressive integrated moving average (SARIMA) model. Based on data from the previous three years, actual and predicted data obtained by these two algorithms were compared and it was found that the prediction errors of Savitzky-Golay filter modification are about 30% smaller then corresponding ones in case of SARIMA model.