Crude Oil Prices Forecast Based on EMD and BP Neural Network

Abstract

The frequent fluctuations in international crude oil prices may affect the stability of the global economy and society. The fluctuation of crude oil prices has nonlinearity, uncertainty and volatility, which bring certain challenges for forecasting crude oil prices. In this paper we use hybrid model with the empirical mode decomposition (EMD) and Back Propagation Neural Network (BPNN) to predict the crude oil prices. To improve the accuracy of prediction, we firstly decompose the crude oil prices data into a series of independent intrinsic mode functions (IMFs) and residual sequences by the empirical mode decomposition method (EMD). Moreover we used BPNN to predict the Brent and WIT crude oil prices respectively. In order to show the effectiveness of the proposed method, we adopt three statistical criteria to evaluate the hybrid method. The empirical results show that EMD-BPNN has higher prediction accuracy than BPNN, the least square support vector regression (LSSVR) and EMD-LSSVR.