A novel method for online real-time forecasting of crude oil price

Abstract

Improving the accuracy of crude oil price forecasting is helpful for stabilizing financial markets and oil import and export trade. However, the extant models rarely focus on the online forecasting and uncertainty of crude oil price. Motivated by these, a new forecasting method is proposed. Firstly, the price sequence is decomposed to get the regular sub-sequences and noise sequence by the improved variational mode decomposition, whose parameters are optimized by particle swarm optimization based on average maximum envelope entropy. Secondly, the proposed hybrid model of point prediction is established by the characteristics of sub-sequences, and the interval prediction model is constructed by combining the point prediction model and Bootstrap sampling. Finally, the proposed point prediction model and interval prediction model are formed into a predictor to realize online real-time prediction. The empirical results show that, compared with other competing models, the proposed model for point prediction improves basically by over 10% on loss functions in the different frequency data, which verifies that it is superior to other models in accuracy and robustness. The coverage rate and fluctuation consistency are almost all more than 70% in the interval prediction results. In a word, the proposed method has high flexibility and accuracy, which can provide referring information for practitioners working with oil.