Technological development has increased the necessity for more accurate predictions that stimulate the application and comparison of modeling techniques and methods of combination. Historically, researchers have believed that combining forecasts from different techniques improves the forecasts, but some studies question whether combining is really the best option. This paper aims to verify whether there is a difference between the accuracy of individual forecasts and that of their combinations by modeling real industrial prediction series. The Box-Jenkins methodology and ANN modeling were used for individual forecasting, whereas the simplified minimum variance and mean arithmetic methods were used for forecast combinations. The performance of the predictions was evaluated by MAPE (Mean Absolute Percentual Error), MSE (Mean Square Error) and MAE (Mean Absolute Error). As the main result, we highlight the predominant frequency at which the predictions obtained by the minimum variance method show superior performance compared to other forecasts.
Forecasting; Accuracy; Combining
