Fuel price forecasting combining wavelet neural network and adaptive differential evolution

Abstract

Once economies are not linearly changing, significant research efforts have been devoted to developing efficient forecasting methods. Artificial neural network (ANN) has been widely applied in forecasting and pattern recognition tasks. Recently, the wavelet neural networks have become a promising tool for nonlinear mapping. In this context, the main of this paper is to forecast the future price for gasoline, diesel, liquid petroleum gas (LGP), liquid natural gas (LNG), and finally sugar cane ethanol. This study differs from previous contributing in literature with three aspects: 1) integration of wavelet analysis and computational intelligence techniques, which are limited in the fuels price forecasting area and are required for assessing the forecasting model for real-life applications; 2) to rank six different neural network structures among the fuels to point the best ones; 3) encourage a discussion about the role of oil price forecasting in wider economic analysis.