Modelo de séries temporais para a previsão de receitas do Fundo Naval

Abstract

This research aims to estimate future revenues from the Naval Fund, a fund of the Brazilian Navy to which public revenues generated within the organization itself are destined, through services provided or other specified inputs. To this end, it used various econometric tools, highlighting univariate and multivariate analyzes between series that make up the fund. The component revenue series were estimated individually, each using the best statistical fit model, with the final modeling having an additive framework between the predicted values for each one, also taking into account the confidence intervals, also predicted. The final product was a flexible perennial and intertemporal mathematical model, which can be adapted year by year for optimization, without loss of functionality, as the Fund's series changes its behavior or there is a change in the eventual relationship with other series, capable of to predict, well in advance, the future flow of revenue. The first quarter and half of the current year were used for comparison purposes between observed revenues and modeling forecasts, seeking values observed in both periods within the additive model's forecast confidence interval.