Prognóstico da série histórica do faturamento da indústria alimentícia brasileira utilizando técnicas de previsão

Abstract

This paper’s objective is to verify which is the best forecasting technique, including the use of the forecasts’ combination to evaluate the prognosis of the Brazilian food industry’s revenues. The historical series of revenues has deterministic trend and seasonality. Thereby, the models chosen to work on were: SARIMA (3,0,0)×(0,1,1)12, SARIMA (4,0,0)×(2,0,0)12 and Holt-Winters Multiplicative. Analyzing the accuracy measures, to perform the series’ forecast it was used the combination of the three models, presented by the methods: Simple Arithmetic Mean, Ordinary Least Squares and Regression of Absolute Minimum Deviation. The results obtained by the forecast were satisfactory, showing that the Brazilian food industry’s revenues will have peaks of growth and decay in the next two years. Therefore, a preparation of the sector is necessary for the period in which a possible decrease in this revenue will occur, as well as dismissal of the workers, since it is the sector that most employs in Brazil.