Crude Oil Price Forecasting Based on Hybrid Deep Learning

Energy prices have always had a significant impact on economic activity as well as on the planning of national energy policies. Under this background, a new decomposition-forecasting-integration hybrid forecasting model is proposed to forecast crude oil prices. Firstly, the primary and secondary decomposition of the original sequence is performed using complementary ensemble empirical modal decomposition and variational modal decomposition, and component reconstruction is performed using random forest and sample entropy. Next, component classification is performed using Lempel-Ziv complexity, and a combined long and short-term memory network-gated recurrent unit forecasting model is used to predict different components. Finally, a convolutional neural network-long and short-term memory network model is used for non-linear reconstruction. The hybrid deep learning model proposed outperforms the other benchmark models proposed in the same dataset, demonstrating the reliability and superiority of this model. At the same time, this paper uses the prediction results and their absolute errors to perform interval prediction based on kernel density estimation, which provides more practical results for prediction.