【摘要】 为帮助烟草企业解决烟用材料采购过程中的成本和风险控制难题，本文提出一种基于自注意力机制的时间卷积网络(Temporal Convolutional Network, TCN)模型，用于科学预测烟用关键辅助材料原纸的上游原材料——纸浆期货的价格。该模块利用卷积模块捕捉长期依赖关系，同时利用自注意力机制学习重点时序段，实现纸浆价格的精准预测。实验结果表明：本文提出的TCN-Attention模型相比LSTM、LSTM-Attention和TCN模型，在纸浆价格预测中的精准度更高。该方法为期货价格预测提供了新思路，也为烟草企业合理制定内衬纸、接装纸的采购价格提供科学参考。更多还原

【Abstract】 To solve the problems of procurement, cost control, and risk management in tobacco enterprises, a Temporal Convolutional Network(TCN) model based on Self-Attention Mechanism was proposed to predict the price of pulp, the upstream raw material for cigarette paper, a key auxiliary material in tobacco manufacturing. The model employs a convolutional module to capture long-term dependencies and uses the self-Attention mechanism to learn the key temporal segments, thus completing the prediction of pulp prices. The experimental results show that the TCN-Attention model proposed in this paper has higher accuracy in pulp price prediction than the LSTM, LSTM-attention, and TCN models. This method provides a new idea for futures price prediction and a scientific reference for tobacco enterprises to reasonably formulate procurement prices for lining paper and tipping paper.更多还原