PRICE FORECASTING FOR FUTURE CONTRACTS ON AGRIBUSINESS THROUGH NEURAL NETWORK AND MULTIVARIATE SPECTRAL ANALYSIS

This study aimed to compare the forecasting results from combining the two models, Multivariate Singular Spectrum Analysis (MSSA) and Artificial Neural Network (ANN), with the results obtained from classical forecasting and neural network models for prices of agricultural future contracts traded on BM&FBOVESPA. The forecasting results of the proposed combination, compared with those obtained from classical forecasting and neural network models showed the best performance for price forecasting. The use of the error measurements and predictive statistical test for the step-ahead confirm this. The research can help market professionals in the development and implementation of risk management policies due to the relevance of price forecasting as a planning tool, in addition to being useful in market behavior analysis in specifying the price trend of future contracts