基于模式匹配与深度学习的国际油价预测研究  

International Oil Price Prediction Based on Pattern Matching and Deep Learning

作　　者：余乐安 雷凯宇 YU Lean;LEI Kaiyu(School of Business,Sichuan University,Chengdu 610065,Sichuan,China;School of Economics and Management,Beijing University of Chemical Technology,Beijing 100029,China)

机构地区：[1]四川大学商学院,四川成都610065 [2]北京化工大学经济管理学院,北京100029

出　　处：《中国石油大学学报（社会科学版）》2023年第5期51-59,共9页Journal of China University of Petroleum (Edition of Social Sciences)

基　　金：中国国家自然科学基金委员会与加拿大魁北克研究基金会合作研究重点项目(72061127002)。

摘　　要：原油是能源安全的基石,故对原油价格的准确预测可以为能源部门的战略规划提供参考。为了提高油价的多期预测精度,基于油价数据的长记忆性特征,构建一种基于模式匹配与深度学习的预测模型。该模型将模式匹配的思想嵌入深度学习模型的训练过程中,并采用3种不同的相似性度量方式对序列之间的相似性进行度量,以增强模型鲁棒性。采用Brent原油价格进行预测实验发现,所提模型在多期预测任务上会显著优于大部分主流预测模型,且模型参数设置对模型的预测效果有影响。Crude oil is the cornerstone of energy security,so accurate prediction of crude oil price can provide a reference for the strategic planning of the energy sector.In order to improve the accuracy of multi-step oil price prediction,based on the long memory characteristics of oil price data,this paper constructs a prediction model based on pattern matching and deep learning.This model embeds the idea of pattern matching into the training process of deep learning model,and uses three different similarity measures to measure the similarity between sequences for enhancing the robustness of the model.The prediction experiment of Brent crude oil price shows that the proposed model is significantly better than most mainstream prediction models for the multi-step prediction tasks,and the setting of model parameters has an impact on the prediction performance of the model.

关 键 词：原油价格预测 相似模式匹配 深度学习 长记忆性 相似性度量 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] F764.1[自动化与计算机技术—控制科学与工程]