人民币汇率预测和风险管理研究
发布时间：2019-07-08 10:38
【摘要】：本文针对汇率问题的复杂性,根据汇率与宏观经济变量的非线性关系和汇率数据自身的非线性特征,以神经网络技术为主要工具,研究了人民币汇率基于购买力平价理论的预测和时间序列预测问题,并建立了一个基于神经网络的人民币汇率风险管理的VaR 模型。 本文应用神经网络考察汇率与物价指数之间的非线性关系。在应用协整技术检验发现购买力平价理论对人民币汇率不成立的情况下,论文建立了一种基于神经网络的非线性协整检验方法,对汇率与物价指数进行了非线性协整检验。并在此基础上建立了预测模型。 论文在分析人民币汇率收益序列长记忆性的基础上,建立了神经网络预测模型。引入方向准确率作为预测模型的一个评价指标,对模型的预测结果进行了统计检验。并考察了计量经济模型的选择标准中的信息准则法,在建立神经网络预测模型过程中的作用。 论文根据神经网络技术中的混合密度网络能够预测数据之间的条件概率密度函数这一特点,对外汇资产组合收益的条件密度函数进行了预测,并在预测的基础上建立了一个VaR 模型。 论文比较系统地分析了人民币汇率的特征,考察了应用神经网络技术进行汇率预测的建模过程和模型选择技术,讨论了对人民币汇率进行有效预测和风险管理的可行性。这些工作,将有助于拓宽对人民币汇率问题的研究思路,为解决人民币汇率预测和风险管理问题提供了一些可供选择的方法。
文内图片：
图片说明：-1资本市场的均衡图中ME表示货币市场的均衡曲线，，它是向上倾斜的
[Abstract]:In view of the complexity of exchange rate problem, according to the nonlinear relationship between exchange rate and macroeconomic variables and the nonlinear characteristics of exchange rate data itself, this paper studies the prediction and time series prediction of RMB exchange rate based on purchasing power parity theory with neural network technology as the main tool, and establishes a VaR model of RMB exchange rate risk management based on neural network. In this paper, neural network is used to investigate the nonlinear relationship between exchange rate and price index. When it is found that purchasing power parity theory is not valid for RMB exchange rate by using cointegration technique, a nonlinear cointegration test method based on neural network is established in this paper, and the nonlinear cointegration test of exchange rate and price index is carried out. On this basis, the prediction model is established. Based on the analysis of the long memory of RMB exchange rate return series, a neural network prediction model is established in this paper. The direction accuracy is introduced as an evaluation index of the prediction model, and the prediction results of the model are statistically tested. The role of the information criterion method in the selection criteria of econometric models in the process of establishing neural network prediction model is also investigated. According to the characteristic that the mixed density network in neural network technology can predict the conditional probability density function between data, this paper forecasts the conditional density function of foreign exchange asset portfolio income, and establishes a VaR model on the basis of prediction. This paper systematically analyzes the characteristics of RMB exchange rate, investigates the modeling process and model selection technology of applying neural network technology to exchange rate prediction, and discusses the feasibility of effective prediction and risk management of RMB exchange rate. These work will help to broaden the research ideas of RMB exchange rate and provide some alternative methods to solve the problems of RMB exchange rate forecasting and risk management.
【学位授予单位】：吉林大学
【学位级别】：博士
【学位授予年份】：2005
【分类号】：F832.6
本文编号：2511530
文内图片：
图片说明：-1资本市场的均衡图中ME表示货币市场的均衡曲线，，它是向上倾斜的
[Abstract]:In view of the complexity of exchange rate problem, according to the nonlinear relationship between exchange rate and macroeconomic variables and the nonlinear characteristics of exchange rate data itself, this paper studies the prediction and time series prediction of RMB exchange rate based on purchasing power parity theory with neural network technology as the main tool, and establishes a VaR model of RMB exchange rate risk management based on neural network. In this paper, neural network is used to investigate the nonlinear relationship between exchange rate and price index. When it is found that purchasing power parity theory is not valid for RMB exchange rate by using cointegration technique, a nonlinear cointegration test method based on neural network is established in this paper, and the nonlinear cointegration test of exchange rate and price index is carried out. On this basis, the prediction model is established. Based on the analysis of the long memory of RMB exchange rate return series, a neural network prediction model is established in this paper. The direction accuracy is introduced as an evaluation index of the prediction model, and the prediction results of the model are statistically tested. The role of the information criterion method in the selection criteria of econometric models in the process of establishing neural network prediction model is also investigated. According to the characteristic that the mixed density network in neural network technology can predict the conditional probability density function between data, this paper forecasts the conditional density function of foreign exchange asset portfolio income, and establishes a VaR model on the basis of prediction. This paper systematically analyzes the characteristics of RMB exchange rate, investigates the modeling process and model selection technology of applying neural network technology to exchange rate prediction, and discusses the feasibility of effective prediction and risk management of RMB exchange rate. These work will help to broaden the research ideas of RMB exchange rate and provide some alternative methods to solve the problems of RMB exchange rate forecasting and risk management.
【学位授予单位】：吉林大学
【学位级别】：博士
【学位授予年份】：2005
【分类号】：F832.6
【引证文献】
相关期刊论文 前1条
1 张晨曦;杨一文;;基于混合密度网络测度股市流动性“周内效应”[J];中国证券期货;2010年09期
相关硕士学位论文 前9条
1 郑东;我国商业银行汇率风险形成机制及其管理研究[D];广东外语外贸大学;2006年
2 张静;开放条件下我国商业银行汇率风险管理问题研究[D];湖南大学;2006年
3 邓雄;我国商业银行汇率风险管理研究[D];西南财经大学;2007年
4 王映乔;基于神经网络的汇率预测及系统设计[D];西南财经大学;2007年
5 周文;基于密度预测的人民币汇率组合预测研究[D];中南大学;2006年
6 姜莹;汇率体制转轨时期我国商业银行外汇风险管理研究[D];南京师范大学;2008年
7 马广元;我国商业银行汇率风险管理研究[D];西南财经大学;2009年
8 卢宁;企业外汇风险防范决策支持系统[D];河南科技大学;2012年
9 侯英超;基于VaR的中国股指期货风险实证研究[D];北方工业大学;2013年
本文编号：2511530
本文链接：https://www.wllwen.com/guanlilunwen/fengxianguanli/2511530.html
