Multi-perspective crude oil price forecasting with a new decomposition-ensemble framework

Crude oil is an important global commodity, and its price fluctuation affects the political and economic security of a country. Therefore, it is necessary to conduct crude oil price forecasting. Based on the forecasting research of multi-source information and decomposition-ensemble, we combine the two into a model and propose a multi-perspective crude oil price forecasting model under a new decomposition-ensemble framework. Specifically, the crude oil price series is decomposed and reconstructed into several modes through variational mode decomposition (VMD) and fuzzy entropy (FE). Further, we screen the effective predictors from structured and unstructured multi-source data using the Granger causality test, and select the optimal input features through random forest - recursive feature elimination (RF-RFE). Finally, each reconstruction mode is individually forecasted on the basis of the selected different input features and the forecasting values obtained are combined and integrated; the final result is obtained from the integrating prediction results through the error evaluation criterion. The West Texas Intermediate (WTI) daily spot price is adopted to validate the performance of our proposed model. The empirical results show that compared with the benchmark models, the proposed model can significantly improve forecasting accuracy.