Pronóstico del precio del petróleo mediante redes neuronales artificiales

Abstract

An artificial neural network model to forecast oil international price is proposed in this work. To develop the model, price data taken from the literature for the WTI reference oil (West Texas Intermediate) traded mainly in New York Mercantile Exchange are used. Four network structures, including the daily price series in the first one, the price series plus the dollar index DXY in the second one, the price series plus the S&P500 index in the third one and the price series plus the DXY and S&P500 indexes in the fourth one are used. Different neural networks configurations are analyzed using a series of six months, where data for five months are used for training patterns and the next month is left for testing the predictive capabilities of the model. The effect of including investors risk aversion using the DXY and S&P500 indexes as alternative input patterns is also analyzed. The results show good performance of the neural networks both during learning and prediction.