A novel hybrid model with two-layer multivariate decomposition for crude oil price forecasting.

Crude oil plays an important role in economic development and political stability, and many scholars have been committed to forecasting its price. However, its influencing factors are complex and diverse, and previous studies have rarely focused on the second multivariate decomposition. Therefore, this study introduces financial market factors and crude oil news as forecasters, and proposes a novel hybrid model with two-layer multivariate decomposition. To verify the performance of the proposed model, an empirical study is performed on weekly West Texas Intermediate (WTI) oil spot price. The results suggest that the second multivariate decomposition for the high-frequency subcomponent can significantly improve the forecasting accuracy, and the forecasting performance of the proposed model outperforms all the benchmark models. • A novel model with two-layer multivariate decomposition and news text is proposed. • Crude oil news variables are constructed based on keywords. • A new means to determine lag order is constructed with statistics and chaos theory. [ABSTRACT FROM AUTHOR]