The Study Of Financial Risk Measurement Based On GARCH-GPD-COOPULA Model

With the financial markets constantly changing, beween of the financial assets show complex relationship.such as a nonlinear. Asymmetric and tail-related features.sometimes the analytical methods basing on linear correlation can’t accurately reflect financial markets related. In reality the financial assets own fat tail.obviously non-normal characteristics and nonlinear.so the traditional VaR calculation method is clearly unreasonable, so we should use a reasonable method to describe the actual distribution and correlation.At the same time the COPULA function method can construct flexible multivariate distributionfunction.It can well description of the rate of return of financial assets and the actual distribution of correlation.we can use it to establish a more effective risk measurement model.Therefore, the use of COPULA function risk value of financial assets has importanttheoretical value and the use of significance.The paper first introduces the GARCH Models and study the distribution of the residuals of the GARCH model, while introducing in both thick-tailed distribution t distribution and GED distribution.then I gives the definition of generalized Pareto distribution and threshold selection methods, after that this paper Comprehensively introduces a COPULA function definition, nature and the five COPULA function, and gives the COPULA function estimation methods, and optimal selection of the COPULA function.On this basis I introduce GARCH-COPULA, the GPD-COPULAand GARCH-GPD-COPULA three models which calculates the value at risk.In the empirical part,Firstly I use the historical simulation method and analysis method to calculate Value at Risk in different quantiles.Then the paper use Monte Carlo simulation method to calculate the GARCH-COPULA. the GPD-COPULA and the GARCH-GPD-COPUL the corresponding value at risk.Finally I use the failure frequency to teste VaR, and compare the five kinds of results in1%2%3%4%5%10%quantile’VaR.The empirical results show that the VaR based on the GARCH-GPD-COPULA method in the sample failure rate is the lowest, it is estimated VaR closest to the truth the VaR.The innovation of this paper is mainly reflected in the following aspects.(1)1Comprehensively summary the definition of COPULA function, classification and estimation methods.(2) When I fit the marginal distribution using GARH model. taking into different residualsare distribution and finally choose the best GARCH model.(3) on the basis of the previous COPULA function, I combinc the GARCH-COPULA and the GPD-COPULA. then proposed GARCH-GPD-COPULA function.