原油期货价格预测对原油开采规划具有重大意义,准确的原油期货价格预测可以实现资源的优化配置和风险的适当规避。在时间卷积网络(Temporal Convolutional Network,TCN)的基础上,利用自适应噪声的完备集合经验模态分解(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise,CEEMDAN)进行原油期货价格数据的特征分解,采用主成分分析(Principal Component Analysis,PCA)对特征分解后的数据进行降维处理,引入自注意力机制(Self-attention,SA)对降维后的数据特征进行注意力分配。结合CEEMDAN、PCA和SA的TCN简记为CEEMDAN-SA-TCN。基于美国西德克萨斯中质原油(West Texas Intermediate,WTI)原油期货价格数据集,利用CEEMDAN-SA-TCN构建原油期货价格预测模型并进行测试。测试结果表明，与线性回归(Linear Regression,LR)、支持向量回归(Support Vector Regression,SVR)、反向传播神经网络(Back Propagation Neural Network,BPNN)、Transformer、Informer、TCN、SA-TCN和CEEMDAN-TCN相比,CEEMDAN-SA-TCN对原油期货价格预测具有更高的准确率,产生的平均绝对误差(MAE)、均方根误差(RMSE)、平均绝对百分比误差(MAPE)的平均值分别为1.642、2.098和1.670。CEEMDAN-SA-TCN可应用于原油期货价格预测中,为原油期货市场的分析与决策提供有力支持。

The prediction of crude oil futures prices is crucial for the planning of crude oil exploitation. Accurate predictions can lead to optimal resource allocation and effective risk mitigation. In this study, we utilized the Temporal Convolutional Network(TCN) in combination with the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN) to decompose the crude oil futures price data. Principal Component Analysis(PCA) was then applied to reduce the dimensionality of the decomposed data, followed by the introduction of Self-attention(SA) to allocate attention to the reduced data features. The combined model integrating CEEMDAN, PCA, and SA with TCN is abbreviated as CEEMDAN-SA-TCN. Using the West Texas Intermediate(WTI) crude oil futures price dataset, a forecasting model was developed and tested. The results show that, compared with Linear Regression(LR), Support Vector Regression(SVR), Back Propagation Neural Network(BPNN), Transformer, Informer, TCN, SA-TCN, and CEEMDAN-TCN, the CEEMDAN-SA-TCN model achieves higher prediction accuracy. The average values of mean absolute error(MAE), root mean squared error(RMSE), and mean absolute percentage error(MAPE) produced by CEEMDAN-SA-TCN were 1.642,2.098, and 1.670 respectively. The proposed CEEMDAN-SA-TCN model in this paper can be applied in prediction of crude oil future prices, providing robust support for the analysis and decision-making in the crude oil futures market.