基于机器学习的铜期货价格预测分析  

Copper futures price forecast based on machine learning

作　　者：沈欣宜 李旭 沈虹[1] SHEN Xinyi;LI Xu;SHEN Hong(School of Business,Yangzhou University,Yangzhou 225127,China)

机构地区：[1]扬州大学商学院,江苏扬州225127

出　　处：《扬州大学学报（自然科学版）》2021年第5期1-7,共7页Journal of Yangzhou University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61803331);江苏省自然科学基金资助项目(BK20170515).

摘　　要：采用支持向量机、MLP(multilayer perceptron)神经网络、LSTM(long short-term memory)神经网络和GRU(gated recurrent unit)神经网络模型,基于基本面信息与市场情绪指标对上海期货交易所铜期货进行多因素价格预测研究.通过选取包括国内外各类经济与金融指标、百度指数等共26个可度量因素,运用相关性分析和主成分分析,共同构建11个特征指标,基于机器学习模型分析其对铜期货价格预测的能力.结果表明:在摆脱对原始交易数据的依赖后,多因素特征指标对沪铜期货价格有较强的长短期预测能力;不同机器学习模型均能得到相似且稳健的预测结果,表明机器学习在期货市场价格预测中具有良好的适用性.This paper adopts support vector machine,MLP neural network,LSTM neural network and GRU neural network model.Based on fundamental information and market sentiment indicators,it conducts multi-factor price forecasting research on the Shanghai Futures Exchange copper futures.This paper selects a total of 26 measurable factors including various domestic,foreign economic and financial indicators,Baidu Index,etc.Through correlation analysis and principal component analysis,11 characteristic indicators are jointly constructed,and their impacts on copper futures are analyzed based on machine learning and deep learning models.The ability to predict prices can be drawn the following conclusions:after getting rid of the dependence on the original trading data,the multi-factor characteristic indicators have strong long-term and short-term forecasting capabilities for Shanghai copper futures prices.Different machine learning models can obtain similar and robust prediction results,indicating that machine learning has a good applicability in futures market price prediction.

关 键 词：机器学习 铜期货价格 SVM模型 LSTM模型 GRU模型 

分 类 号：F224.13[经济管理—国民经济]