<![CDATA[Given the critical importance of crude oil prices in the global economy, this study focuses on forecasting WTI and Brent crude oil futures prices. To incorporate both price series as independent variables in our predictive model, we employed Vector Autoregression (VAR) to analyze their interdependence. Results from impulse response analysis, variance decomposition, and Granger causality tests revealed a significant spillover effect between the two crude oil futures prices over the past 12 lagged periods. Consequently, the lagged data from the past 12 periods of both price series were utilized as independent variables for the deep learning framework. We developed a hybrid model integrating Empirical Mode Decomposition (EMD), Autoregressive Moving Average (ARMA), and Bidirectional Long Short-Term Memory (BiLSTM) to predict WTI and Brent prices. Our findings demonstrate that predictive performance improves with increased model complexity. Specifically, the hybrid model (EMD+ARMA+BiLSTM) outperforms both the EMD+BiLSTM model and the standalone BiLSTM model. For WTI, the hybrid model achieved a Mean Squared Error (MSE) of 57.77, Mean Absolute Error (MAE) of 5.66, and Mean Absolute Percentage Error (MAPE) of 8.78%. For Brent, the corresponding metrics were 62.15, 5.87, and10.40%, respectively. These results validate the robustness and rationality of our hybrid model, offering a reliable methodology for crude oil price prediction and providing valuable insights for researchers and practitioners in the energy markets.]]>
Copyrights © 2026
