Data for: A novel hybrid method for crude oil price forecasting
Description
Abstract of associated article: Forecasting crude oil price is a challenging task. Given the nonlinear and time-varying characteristics of international crude oil prices, we propose a novel hybrid method to forecast crude oil prices. First, we use the ensemble empirical mode decomposition (EEMD) method to decompose international crude oil price into a series of independent intrinsic mode functions (IMFs) and the residual term. Then, the least square support vector machine together with the particle swarm optimization (LSSVM–PSO) method and the generalized autoregressive conditional heteroskedasticity (GARCH) model are developed to forecast the nonlinear and time-varying components of crude oil prices, respectively. Next, the forecasted crude oil prices of each component are summed as the final forecasted results of crude oil prices. The results show that, the newly proposed hybrid method has a strong forecasting capability for crude oil prices, due to its excellent performance in adaptation to the random sample selection, data frequency and structural breaks in samples. Furthermore, the comparison results also indicate that the new method proves superior in forecasting accuracy to those well-recognized methods for crude oil price forecasting.