Previsão da demanda de exames de direção veicular no estado do Rio Grande do Sul

Abstract

The objective of this work is to model and to foresee the weekly demand of transportation direction exams that were accomplished in Rio Grande do Sul. The data were contained in 122 (hundred and twenty-two) weekly observations, collected of reports generated by FATEC/UFSM, in the period of July 1st, 2003 to October 28th, 2005. The used methodology was the Time Series Analysis, through the models ARIMA(p,d,q), with 95% level (ninety five percent) of trust. The behavior of the original series was analyzed and verified the nonstationary of the data. For stationary the series the transformation was accomplished by the differences method. After that the autocorrelation functions and partial autocorrelation were verified. The parameters were esteemed for the suitable models. The models ARIMA(1,1,1) and IMA(1,1) they were the most appropriate for the SBIC criterion; for the AIC criterion, the most appropriate model was ARIMA(1,1,1); however, it it considers the principle of the parsimony that the more simple model is, the better it will be, the model IMA(1,1) was chosen. The models ARIMA(1,1,1) and IMA(1,1) they supplied good forecasts, however if it compares those models, starting from the subjective analysis accomplished in the FATEC Scheduling Section of Trabalhando pela Vida!, the model that presented the smallest mistakes and it is closer of the reality of the weekly transportation direction exams of RS is ARIMA(1,1,1). The equations of the models were adjusted and accomplished the forecasts of the demands of transportation direction exams for 23 (twentythree) horizon steps ahead, through the statistical software Minitab 14.2.