Previsão dos preços de commodities agrícolas brasileiras no mercado futuro utilizando redes neurais artificiais

Abstract

The commodities market is characterized by being dynamic and complex. Futures market prices are traded on stock exchanges and short-term variables such as exchange rate changes, future payments, government reports, world stocks, among others. Considering the instability of the commodities market, this work aims to analyze the construction of artificial neural networks for forecasting agricultural commodity prices in the future market, with the objective of evaluating the networks as a strategic measure in the process of making decision. It is an empirical research, of descriptive character, with quantitative approach. The method used was the survey, using data base results in the database of Cepea. As networks were developed with indicators as "optimal" for all commodities, which demonstrated a high predictability of artificial neural networks. Through the construction of monthly commodity scenarios, it was verified that the selected networks are sensitive to the variations of rise and fall in prices. Thus, the RNAs have proved to be an important tool that can successfully aid the search in this market, through the simulation of the behavior of prices.