A hybrid system with optimized decomposition on random deep learning model for crude oil futures forecasting

The violent fluctuations in crude oil prices have a significant influence on the global economy. Therefore, it is crucial to establish a precise and reliable model for forecasting crude oil prices. To enhance the forecasting accuracy of crude oil futures prices, this study introduces an innovative hybrid approach. The proposed approach integrates optimized variational mode decomposition, an attention mechanism and the stochastic error correction algorithm into the deep bidirectional long short term memory network. First, a novel decomposition approach is introduced that incorporates the concept of cross fuzzy entropy into variational mode decomposition. Second, an attention mechanism is employed to improve the effectiveness of the forecasting framework. Third, the stochastic strength formula is employed to correct errors during the model training phase, consequently strengthening the acquisition of efficient characteristics through historical data weight adjustment. Finally, two sets of crude oil futures sequences are applied for empirical predictions. The proposed model exhibits outstanding performance by studying the results of multiple evaluations and a novel error measurement. Compared with those of the other models, the final empirical results demonstrate that the proposed forecasting model can enhance the accuracy of crude oil futures forecasts.