Análise do comportamento futuro do preço de compra do cimento Portland CP IV

Abstract

The main purpose of this research is to predict the Portland CP IV cement purchase price for a supply materials store, by means of autoregressive integrated moving average (ARIMA) forecasting models and forecast combinations using the selected ARIMA models. Found the best forecast values by ARIMA or forecast combination, these will be investigated by residual individual control charts derivate from the forecasting accomplished. The control chart will indicate the moment of high and drop cement purchase price, in order to assist the decision making manager to maintain or increase the inventories. The models found were ARIMA (0,1,4)C, ARIMA (0,1,1) C and ARIMA (4,1,0) C show the best forecast performance. Better than the individual ARIMA models, the forecast combination using OLS provide MAPE forecast statistic equal to 2,882%, the AIC´s combination proposed forecasts show similar results to OLS method. To test the difference between OLS and AIC´s method the nonparametric Wilcoxon test was made proving that both methodologies are statistically equal with p-value = 6.78%.