Suporte a decisão no setor sucroalcooleiro

Abstract

The sugar and alcohol sector is one of the largest agricultural sectors in Brazil. Each harvest millions of liters of ethanol and thousands of tons of sugar are imported worldwide.Despite the size of the sector, there are several problems that haunt the sugarcane producer. One is the drop in production causing sugar and ethanol production stops.This paper aims to carry out a comparative study of time series forecasting methodsin historical sugarcane production data, together with the construction of operation al indicators to aid in decision making. The database was taken from the quarterly results published by São Martinho for its investors. São Martinho is a publicly traded companyand one of the largest sugar, alcohol and energy production plants in Brazil. The R language was used to carry out the study. The experiments of this work used the predictive model SARIMA, for its almost unanimity in the forecast of agricultural yields.RMSE, ME, and MAE. In the development of the operational indicators, the waste distribution function of the SARIMA model defined along with the forecasts of the modelitself was used.At the end of all the work, the best SARIMA model was obtained for the quarterly sugarcane production data together with the indicators of fall in production: probability off all in production by 30 % and probability of fall in production below quarterly average production.