Forecasting Crude Oil Price Based on EMD-Wavelet-GARCH Model

Abstract

Aimed at solving the problem of international crude oil price forecasting, a decomposition combination forecasting method is proposed based on empirical mode decomposition (EMD)wavelet threshold denoise method, Generalized auto regressive conditional heteroskedastic (GARCH) and decision tree. EMD is used to decompose the international crude oil price series into several IMFs with different frequencies. According to the T-test method, the IMF components are simply grouped into high-frequency, low-frequency and trend series. High-frequency series are influenced by the market fluctuations, low-frequencies series are influenced by major events, and the trend series show the overall price volatility in the oil market. The high-frequency sequence is predicted by the time series GARCH model. The low-frequency sequence is predicted by the decision tree model. The predicted high-frequency sequence and the predicted low-frequency sequence are denoised by wavelet denoise. And the trend sequence is added to obtain the simulated crude oil price. The average error and root square error show that the model has high predictability and is suitable for the analysis and prediction of crude oil market prices.