基于混合深度学习的原油价格预测

Crude Oil Price Forecasting Based on Hybrid Deep Learning

能源价格对经济活动以及国家能源政策的规划一直影响重大。基于这一背景,提出了一种新的分解-预测-集成混合预测模型来预测原油价格。首先运用互补集成经验模态分解和变分模态分解对原始序列进行一次和二次分解,运用随机森林和样本熵进行分量重构;接着,利用Lempel-Ziv复杂度进行分量分类,运用长短期记忆网络-门控循环单元组合预测模型,预测不同分量;最后利用卷积神经网络-长短期记忆网络模型进行非线性重构。所构建的混合深度学习模型在同一数据集中的性能优于所提出的其他基准模型,证明了所提出模型的可靠性和优越性。同时,基于预测结果及其误差进行的基于核密度估计的区间预测结果,为原油价格预测提供了更具有实用性的依据。

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.