基于HMM的VaR风险度量及其实证分析

发布时间：2018-06-24 05:38

本文选题：隐马尔可夫模型 + VaR风险价值　；参考：《合肥工业大学》2013年硕士论文

【摘要】：新世纪特别是2008年世界金融危机爆发以来，国际、国内金融市场都发生了深刻的变革，金融市场风险明显增大，分析金融波动特征和度量金融市场风险对投资与监管都具有重要的意义。VaR(Value-at-Risk)方法以其高度的综合概括能力，为投资者提供了一个直观、全面的风险量化指标，，目前已成为世界主流的风险测度方法。本文主要针对金融数据所呈现的尖峰厚尾、波动持续、结构转变等特征，探讨了金融市场风险度量的VaR模型的改进方法与应用。首先对本文的研究背景及意义、VaR模型在国内外的研究概况进行概述，并提出了本文的主要内容。然后介绍了隐马尔可夫模型的基本概念与算法，指出其在异常状态识别上的应用。紧接着阐述了VaR的基本原理与计算方法，指出常用的ARCH模型族估算波动率方法的不足，在此基础上提出了本文模型：HMM-ARMA-GARCH模型，用隐马尔可夫模型的状态变量来描述金融市场的正常波动状态与异常波动状态，同时不可观测的状态变量能够对波动的集聚现象给出很好的解释。对不同的状态数据分别建立ARMA-GARCH模型来估算波动率，同时给出VaR的具体计算方法。最后对上证企债指数进行实证分析，采用Kupiec失败频率检验法对VaR的准确性进行检验，并与传统的ARMA-GARCH模型的估算效果进行比较。实证结果表明基于本文模型的VaR计算方法具有较好的估计效果，且能够有效的降低GARCH模型高估波动持续性的现象。
[Abstract]:In the new century, especially since the outbreak of the world financial crisis in 2008, profound changes have taken place in the international and domestic financial markets, and the risks in the financial markets have obviously increased. It is important to analyze the characteristics of financial volatility and to measure financial market risk for investment and supervision. The VaR (Value-at-Risk) method, with its high comprehensive generalization ability, provides an intuitive and comprehensive risk quantification index for investors. At present, it has become the mainstream risk measurement method in the world. Aiming at the characteristics of financial data, such as peak and thick tail, persistent fluctuation and structural transformation, this paper discusses the improved method and application of VaR model for financial market risk measurement. Firstly, the research background and significance of this paper are summarized, and the main contents of this paper are put forward. Then the basic concept and algorithm of hidden Markov model are introduced, and its application in abnormal state recognition is pointed out. Then, the basic principle and calculation method of VaR are expounded, and the shortcomings of the common arch model family are pointed out. On this basis, the model of this paper is proposed, which is the: HMM-ARMA-GARCH model. The state variables of hidden Markov model are used to describe the normal volatility and abnormal volatility in the financial market. At the same time, the unobservable state variables can give a good explanation for the agglomeration of volatility. The ARMA-GARCH model is established for different state data to estimate volatility, and the specific calculation method of VaR is given. Finally, the empirical analysis on the debt index of Shanghai stock market is carried out, and the accuracy of VaR is tested by using the Kupiec failure frequency test method, and compared with the estimation effect of the traditional ARMA-GARCH model. The empirical results show that the VaR calculation method based on this model has better estimation effect and can effectively reduce the phenomenon of overestimating volatility persistence in GARCH model.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F830.91;F224;O211.62

【参考文献】