Statistical Analysis Of Markov Regime-switching Model And Its Applied Research In Finance

The Markov Regime Switching model is a dynamic time series model of the popular,uncertain switching it can not completely predictable,as a random variable in time would be random and continuity.With the continuous development of Markov Regime Switching model,it is widely used in the field of economy and finance,combining with the traditional time series model.In a large number of studies,it is found that the logarithmic returns of financial time series are characterized by fluctuating agglomeration,structural instability and post-peak tail,and traditional time series can not describe the characteristics of financial time series well.In order to solve this problem,the Markov Regime switching model is combined with the traditional time series model,and the models of - and - are proposed to solve the problem of structural mutation.However,since the residual sequence obeys different distributions,it extends from a normal distribution to a biased distribution such as distribution and generalized error distribution,which greatly improves the fitting effect of the model.This paper choose from January 4,2005 to September 29,2016,the Shanghai index,and on November 2,2009 closing to the exchange rate on September 29 th,2016 closing price as the research object.Based on were used respectively to measure the estimate of model and estimation methods under different distribution based on - model to data modeling,and the comparative analysis of the model parameters.On this basis,the - model under different distributions is used to predict the out-of-sample data,and the results are analyzed.At the same time,the model under the Gauss Mixture Model is used to test the state number of the data.The empirical results show that:(1)The parameters of model are estimated by using Monte Carlo( )estimation method,and the data is analyzed by constructing the coefficient algorithm,and compared with - - model By comparing the parameters,it is found that the - - model is better able to describe the data.(2)The study on the sustainability of the data,using estimation method based on different distributions under the - model,the data outside the sample prediction, distribution of the predicted value closer to the true value,and the fitting effect and the volatility description is better.(3)The - model of the two states is tested by hypothesis using the algorithm of the Gaussian Mixture Model.By estimating the log-likelihood estimates,the two states are divided into the best fitting states.