基于GARCH和LSTM神经网络混合模型的人民币汇率预测研究  

Prediction of RMB Exchange Rate Based on GARCH and LSTM Neural Network Hybrid Model

作　　者：郑扬飞 张青龙[1] Yangfei Zheng;Qinglong Zhang(School of Management,University of Shanghai for Science and Technology,Shanghai,200093,China)

机构地区：[1]上海理工大学管理学院,上海200093

出　　处：《管理科学与研究（中英文版）》2022年第9期121-127,共7页Management Science and Research

摘　　要：本文选取了深度学习算法中对时间序列预测表现良好的长短期记忆模型(Long Short Term Memory,LSTM)与GARCH模型相结合,对2017年1月至2021年12月美元兑人民币和欧元兑人民币汇率进行预测,以融合新型的深度学习模型和传统的金融时间序列预测模型各自优势来提高人民币汇率预测的准确性。经实证发现,GARCH模型、LSTM模型、GARCH和LSTM的混合模型预测的准确性依次提高,混合模型能较好地提升预测能力。In this paper,the Long Short Term Memory model with good performance in time series prediction in deep learning algorithm was selected to combine with GARCH model to predict the exchange rate of USD/RMB and EUR/RMB from January 2017 to December 2021.The accuracy of RMB exchange rate prediction is improved by combining the advantages of the new deep learning model and the traditional financial time series forecasting model.Empirical results show that the prediction accuracy of GARCH model,LSTM model,and mixed model of GARCH and LSTM are improved successively,and the mixed model can better improve the prediction ability.

关 键 词：人民币汇率预测 GARCH模型 LSTM神经网络 

分 类 号：F83[经济管理—金融学]