首页 > 资料专栏 > 论文 > 营销论文 > 市场管理论文 > 基于深度学习的外汇市场预测与风险测度研究_MBA毕业论文DOC

基于深度学习的外汇市场预测与风险测度研究_MBA毕业论文DOC

汇市明灯
V 实名认证
内容提供者
资料大小:3684KB(压缩后)
文档格式:DOC
资料语言:中文版/英文版/日文版
解压密码:m448
更新时间:2018/12/30(发布于广东)

类型:金牌资料
积分:--
推荐:升级会员

   点此下载 ==>> 点击下载文档


文本描述
摘要
随着全球一体化的加速发展与国际贸易范围的不断扩大,国际金融市
场上涉及贸易往来的资金规模不断增大,互联网与高性能计算机技术的发
展也促进了跨国贸易的发展,它们不仅使得国际间的贸易往来越来越频
繁,而且使得形式也越来越多样化,汇率作为国际间经济贸易活动的重要
纽带之一,发挥着越来越重要的作用

在人民币汇率改革之后,汇率形成机制的市场化特征不断增强,人民
币汇率变动更加剧烈。近些年来,外汇市场受到东南亚金融危机,美国次
贷危机以及欧债危机的影响,国际上主要货币的汇率走势依旧复杂难测,
外汇市场风险不断增加,汇率预测与风险度量成为汇率研宄领域的重要问
题之一。同时,加强外汇市场汇率预测的准确度以及对汇率风险的管理对
我国跨国企业经营活动十分重要

深度学习模型自提出以来,在图像识别、语音识别、分类、数据挖掘
等方面都有着较好的表现,在处理非线性问题方面有着较高的性能。汇率
的历史数据序列呈现显著的非线性特征。目前的经典模型都无法对这其中
的内在关系进行较好的描述以及特征提取,本文将深度学习模型引入到金
融预测与风险度量领域,对汇率市场历史数据中的复杂非线性特征进行建
模分析

针对汇率走势的精确预测与汇率风险的精确估计问题,本文提出基
I 于深度学习理论的汇率预测与风险度量模型。首先,本文介绍了深度学习
模型近年来在图像识别、语音识别等多个领域的研宄现状,然后提出一种
新的基于深度学习模型的外汇市场汇率预测模型。之后基于VaR风险度
量理论,提出一种新的基于深度学习理论的集成VaR风险度量模型。该
模型将深度学习模型与ARMA-GARCH模型相结合,构建了 VaR估计模
型。并且釆用7个主流外汇市场的汇率数据进行实证研宄,实证结果表明
本研究中所提出的预测与风险度量模型具有较好的精确性以及可靠性

关键字:汇率预测,汇率风险度量,Deep Belief Network (DBN)模型,
Long Short Term Mmeory (LSTM)模型,Value at Risk (VaR)模型
II ABSTRACT
RESEARCH ON FORECASTING AND VALUE-AT-RISK
MEASUREMENT IN FOREIGN EXCHANGE MARKETS
BASED ON DEEP LEANING
ABSTRACT
With the accelerated development of global integration and the
continuous expansion of international trade, the scope of the funds for trade in
international financial markets is growing. The development of Internet and
high-performance computer technology have also promoted the development
of cross-border trade activities, making International trade is becoming more
and more frequent, and the forms are becoming more and more diversified.
The exchange rate plays an increasingly important role as one of the important
links of international economics and cross-border trade activities.
After the reform of the RMB exchange rate, the market-oriented
characteristics of the exchange rate formation mechanism are increasing, and
the exchange rate of the RMB changes more violently. In recent years,
Southeast asian financial crisis, subprime crisis, the euro crisis have a serious
negative impact on the foreign exchange market, the international exchange
rate of major currencies is still complex and unpredictable, the risk of foreign
exchange market continues to increase, exchange rate forecast and risk
measurement into exchange rate research field One of the important research .
At the same time, it is very important to strengthen the accuracy of exchange
in
北京化工大学硕士学位论文
rate forecast in foreign exchange market and the risk management of exchange
rate to multinational enterprises in China.
The deep learning model has a good performance in image recognition,
speech recognition, classification, data mining and so on, and has high
performance in dealing with nonlinear problems. The historical data of the
exchange rate also has a very complex linear and non-linear relationship, the
current classic models can not be a good description of which the relationship
between the intrinsic description and feature extraction, Therefore, it is a new
attempt to describe the complex intrinsic function of exchange rate market
historical data by applying the depth learning model to the field of financial
forecasting.
Aiming at the problem of accurate forecasting of exchange rate
movements and more accurate estimation of exchange rate risk, this paper
proposes a model of exchange rate forecast and risk measurement based on
deep learning theory. Firstly, this paper introduces the research status of the
deep learning model in image recognition and speech recognition in recent
years, and then proposes a new exchange rate forecasting model in foreign
exchange market based on the deep learning model. Then, a new risk
management model of integrated VaR is proposed based on the theory of risk
measurement. This model combines the depth learning model with the
ARMA-GARCH model to construct the VaR model. And exchange rate data
of seven mainstream foreign exchange markets is used in the empirical
study,the model proposed in this study has good accuracy and reliability.
KEYWORDS: Exchange rate forecast, Exchange rate risk measurement,
Deep belief network model (DBN) ,Long short term memory model
(LSTM) ,Value at Risk model ( VaR)
IV
目录
m-M i
i.i研宄背景 i
1.2研宄目的及意义 2
1.3国内外研宄综述 3
1.3.1外汇市场预测的研宄综述1.3.2外汇市场风险度量研宄综述1_3.3 VaR风险度量方法的研宄综述1.3.4集成算法的研宄综述
1〇
1.3.5深度学习方法的研宄综述1.4本文研宄内容以及创新点i
2.1祌经网络 18
2.1.1神经元 18
2.1.2激活函数 18
2.2深度学习模型 20
2.2.1 深度置信网络(Deep Belief Networks,DBNs)2.2.2深度多层感知器
23
2.2.3 循环神经网络(Recurrent Neural Networks,RNNs)
24
2.3优化算法 28
2.3.1共轭梯度法 28
2.3.2随机梯度下降算法
29
2.4本章小结 30
第三章基于深度学习的外汇市场汇率预测模型
31
3.1 弓 IW 31
3.2基于深度学习的外汇市场汇率预测模型
31
3.2.1数据处理 31
V
北京化工大学硕士学位论文
3.2.2预测模型结确定
32
3.2.3基于深度学习的集成汇率预测
33
3.3实证研宄 33
3.3.1实验数据 33
3.3-2实证结果分析
34
3.4本章小结 36
第四章基于深度学习的集成VaR风险测度模型
37
4.1弓丨胃 37
4.2 VaR风险度量理论
37
4.3基于深度学习的VaR风险测度模型
39
4.3.1数据处理 39
4.3.2分风险估计 40
4.3.3总风险集成 41
4.4实证研究
42
4.4.1实验数据 42
4.4.2评价指标 43
4.4.3实证结果分析
43
4.5本章小结 46
第五章麟和展M
47
参考文献 49
致谢 59
作者和导师简介
61
VI Contents
Contents
Charpter 1 Introduction1.1 The backgroud on reseaarch1.2 Research purposes and significance1.3 Literature review at home and aborad1.3.1 Literature review of forecasting in foreign exchange market1.3.2 Literatxire review of risk management in foreign exchange market1.3.3 Literature review of VaR1.3.4 Literature review of ensemble algorithm1.3.5 Literature review of deep learning1.4 he innovations ofthis thesisCharpter 2 Deep learning theory2.1 Neural networks2.1.1 Neuronal structure2.1.2 Activate function
;2.2 Models based on deep learning2.2.1 Deep Belief Networks(DBNs)2.2.2 Deep neural networks (DNN)
23
2.2.3 Recurrent Neural Networks (RNNs)
24
2.3 Optimization algorithm
28
2.3.1 Conjugate Gradient Algorithm (CG)
28
2.3.2 Stochastic Grandient Descent Algorithm (SGD)
29
2.4 Summary 30
Charpter 3 Forecasting of exchange rate in foreign exchange mark
et based on Deep learning model
31
。。。以上简介无排版格式,详细内容请下载查看