Forecasting crude oil prices based on variational mode decomposition and random sparse Bayesian learning

• Published in: Applied soft computing 2021-12, Vol.113, p.108032, Article 108032
• Main Authors: Li, Taiyong, Qian, Zijie, Deng, Wu, Zhang, Duzhong, Lu, Huihui, Wang, Shuheng
• Format: Article
• Language: English
• Subjects: Crude oil price forecasting; Decomposition and ensemble; Price prediction; Sparse Bayesian learning (SBL); Variational mode decomposition (VMD)
• ISSN: 1568-4946; 1872-9681
• DOI: 10.1016/j.asoc.2021.108032

Accurately forecasting crude oil prices has drawn much attention from researchers, investors, producers, and consumers. However, the complexity of crude oil prices makes it a very challenging task. To this end, this paper presents a novel scheme by integrating variational mode decomposition (VMD) and random sparse Bayesian learning (RSBL, SBL-based prediction with random lags and random samples), namely VMD-RSBL, for the forecasting task. The proposed VMD-RSBL contains three stages. First, crude oil price series is decomposed into a couple of components by VMD. The decomposed components exhibit simpler characteristics than the raw prices and hence are easy to forecast. Second, RSBL is employed to predict each component individually. Specifically, for each component, the proposed scheme builds a group of predictors with SBL on different subsets of samples (random samples) and random lags, and then the average of all the predictors is taken as the forecasting result of the individual component. At last, the forecasting results of all the components are added as the final forecasting prices. We perform extensive experiments, and the results demonstrate that the proposed VMD-RSBL significantly outperforms many state-of-the-art schemes in terms of several evaluation indicators. •Random sparse Bayesian learning (RSBL) is proposed for individual prediction.•Variational mode decomposition and RSBL are integrated to forecast crude oil prices.•The proposed approach is superior to the state-of-the-art methods in terms of prediction accuracy.