基于深度学习长短期记忆神经网络的有色金属期货市场预测研究被引量：8

Prediction of non-ferrous metals futures market based on deep learning LSTM neural network

下载PDF

导出

摘要为提高金融资产预测能力,该文采用深度学习长短期记忆(LSTM)神经网络模型对上海期货交易所(SHFE)和伦敦金属期货交易所(LME)的铝、铜、镍、铅、锡和锌6种有色金属期货价格分别进行长、短期预测,与传统机器学习多层感知器(MLP)模型以及线性自回归移动平均(ARIMA)模型进行对比研究。数据源于Wind数据库和国际货币基金组织(IMF)数据库。使用Python深度学习软件模拟预测有色金属期货价格,结果显示:有色金属期货市场长期预测中,LSTM模型的预测表现略逊于ARIMA模型,MLP模型预测效果不理想;短期预测中,LSTM模型的预测结果和ARIMA模型相近,均优于MLP模型;LSTM模型与MLP模型相比,模型的稳定性和预测的精确度都更加出色。LSTM模型可作为ARIMA模型的最优替代之一。 To improve the predictive ability of financial assets,the deep learning long short-term memory(LSTM)neural network model is used to forecast the futures prices of six kinds of non-ferrous metals(aluminum,copper,nickel,lead,tin and zinc)in Shanghai Futures Exchange(SHFE)and London Metal Futures Exchange(LME).The LSTM is compared with the traditional machine learning multi-layer perceptron(MLP)model and linear autoregressive integrated moving average(ARIMA)model.The data is from the Wind database and the International Monetary Fund(IMF)database.Python deep learning software is used to simulate and predict the futures price of non-ferrous metals.The results show that in the long-term forecast of the non-ferrous metals futures market,the forecasting performance of the ARIMA model is slightly better than that of the LSTM model,and the MLP forecasting effect is not ideal;in the short-term prediction,the prediction results of the ARIMA model are similar to the LSTM model,and both are better than the MLP model;compared with the MLP model,the LSTM model has better model stability and prediction accuracy.The LSTM model can be used as one of the best alternatives to the ARIMA model.

作者沈虹李旭潘琪 Shen Hong;Li Xu;Pan Qi(School of Business,Yangzhou University,Yangzhou 225127,China)

机构地区扬州大学商学院

出处《南京理工大学学报》 CAS CSCD 北大核心 2021年第3期366-374,共9页 Journal of Nanjing University of Science and Technology

基金国家自然科学基金(61803331) 江苏省自然科学基金(BK20170515) 扬州大学科技创新基金(X20200892) 扬州大学商学院创新计划(SX2020025)。

关键词深度学习长短期记忆模型神经网络多层感知器模型自回归移动平均模型有色金属期货市场价格预测 deep learning long short-term memory model neural network multi-layer perceptron model autoregressive integrated moving average model non-ferrous metals futures markets prices prediction

分类号 F222 [经济管理—国民经济]