Modelagem de séries temporais para projeção do volume de concessão de empréstimo consignado para aposentados e pensionistas do INSS

Abstract

The present work aims to use statistical time series models to project, with a certified degree of confidence, using the R software, the volume of payroll loans granted to INSS retirees and pensioners, taking into account the degree of fidelity of this modality and this niche of credit takers within various credit granting modalities. The series followed comprised the period from May 2011 to December 2021, and after statistical analysis of the series and through the Box and Jenkins method, different models were adjusted, observing the comparison of the adjustment statistics and the forecast error statistics to find the best model that accompanies the series. The great variability in the final period of the series hindered the search for the best model. The results obtained indicate that the use of exogenous variables can contribute to a better forecast of the series, being the SARIMAX model with seasonality in time t=12, which presented the best projection.