Empirical Research On Stock Index Of Value At Risk On Bayesian MCMC Algorithm

With the rapid development of financial market globalization,financial products are facing many risks because of the changes in financial products,national policies and information technology innovation.Financial risk refers to the possibility of actual returns deviating from expected returns due to factors such as the uncertainty of economic activities,changes in market environment,and errors in decision-making.The Va R model is a method of quantifying risk,which uses loss as an indicator of monitoring.It is of great significance to study a solution method that conforms to the characteristics of the model,reducing the estimation error,scientifically assessing the expected loss of risk,and avoiding making mistakes due to overestimation or underestimation of risk.The core of the VaR model is to estimate the fluctuation of the return rate.The GARCH cluster model is used to describe the volatility and the Va R model based on it is to measure the risk value.The GARCH cluster model can reflect the distribution characteristics of data spikes and tails.At the same time,the GJR-GARCH model shows the asymmetric effect of different messages on volatility.Under the classical method,the maximum likelihood estimation method is used to solve unknown parameters.In this paper,Bayesian framework was introduced to complete the statistical inference.The MCMC simulation of each parameter in the model is carried out by Griddy-Gibbs sampling and Metropolis-Hastings algorithm to provide theoretical support for empirical research.During the empirical research process,the Shanghai Stock Index data is divided into two parts.The first part is used as a training to build model,the latter part is used as a test sample to test model.The prediction results of Va R calculation methods based on historical simulation,Monte Carlo simulation and GARCH cluster model are tested by Kupiec method.The empirical results show that the effect of Va R prediction based on the GARCH cluster model is significantly better than the traditional methods such as historical simulation and Monte Carlo simulation.Furthermore,the GARCH model based on the Bayesian framework is superior to the maximum likelihood estimation result.The MCMC method is more flexible and reliable in the statistical inference process.The Va R model obtained by Bayesian MCMC method is a new way and tool to control internal risk and measure risks for corporate and regulatory authorities.