A Novel Multivariate Decomposition Ensemble Model with News Text for Crude Oil Price Forecasting

Forecasting crude oil prices is crucial for every individual and even the entire country. Previous studies have encountered difficulties in forecasting highly nonlinear crude oil prices, especially when conflicts, wars, and other irregular events occur. In light of this, this study proposes a novel multivariate decomposition ensemble model with news text for crude oil price forecasting, which mainly consists of four steps. First, data fusion of multivariate forecasters is performed. Second, the crude oil price and its forecasters are decomposed and reconstructed using multivariate empirical mode decomposition (MEMD) and sample entropy (SE), respectively. Thereafter, the effective forecasters are screened from the reconstruction subcomponents of forecasters through the Granger causality test. Finally, the crude oil price is forecasted using a hybrid forecasting technique, and the validity of the proposed model is evaluated from different perspectives. The empirical results indicate that the proposed model achieves excellent performance in forecasting the West Texas Intermediate weekly spot price