Pronóstico de precios de petróleo: una comparación entre modelos garch y redes neuronales diferenciales

The aim of this paper is to show the advantages of the use of neural networks differentials (rnd) as an efficient alternative method in calculating the forecasts of future prices of financial assets, for which a comparison is made with models of the garch family, to carry out the forecast of future closing price of crude oil barrels, types West Texas International and Brent. The results shows that the use of rnd has essentially the same accuracy as the values obtained with the tgarch (1,1) model and are superior to those obtained by the garch (1,1) model to calculate price forecasts barrels of crudes Brent and wti respectively during the period of description, from January 2, 2013 to February 24, 2015 and the forecast period from February 25 to March 10, 2015. However, the effort made to obtain such results with the family of garch models is significantly higher than when using the rnd, this supports the proposal to use the rnd as a reliable alternative method in the analysis of time series.