Ajustamento de um modelo de previsão por meio da técnica autorregressiva de médias móveis e entrada exógena para a produção brasileira de biodiesel

Abstract

The objective of this research was to analyze the Brazilian biodiesel production considering the influence of macroeconomic variables, as well as adjusting a forecast model through ARMAX modeling. A descriptive analysis of the variables was performed, as well as the causality analysis using the Granger test. Finally, the Brazilian biodiesel production was estimated by ARMAX modeling. It was observed the existence of direct relations between Biodiesel Production with Soybean Oil Price, Soybean Price, Oil Production, Natural Gas Production and GDP. The adjusted model that best describes the production of biodiesel was a SARIMAX (0,1,2) (1,0,0) 12 Δ (Price of Soybean Oil) t-3 Δ (Production of Oil) t-5 Δ ( GDP) t-2 Δ (Natural Gas Production) t-2; t-5. This model was able to adequately capture the behavior of the analyzed variable, from the influence of external variables, since the original values from July to December 2017 are within the confidence interval. In addition, the values of the model accuracy statistics, MAE (22424,74), MAPE (5,6992%) and U-Theil (0,026921) were adequate. Thus, the results of the present research contribute to the study area on biodiesel, because through the modeling it was possible to obtain an accurate forecast of the Brazilian biodiesel production, and in this sense the implementation of advanced techniques to predict both production and the consumption of biodiesel or other variables that may influence the behavior of biofuels can be considered as effective alternatives and source of information for decision making regarding the planning of the biofuels sector.