Identification Analysis And Application Of Change Points Of Multiple Time Series Models

The change point theory is a classic branch of statistics.The recognition of change point uses statistical methods to estimate the position of the change point.Firstly,the Bayesian theory and the recognition theory of various time series models are introduced.Then,based on Bayesian theory,the change point recognition methods are explained and used to complete the programming of the algorithm for the homogeneous variance and the heteroscedastic time series model.The statistical software Matlab is used to simulate various time series models to verify the effectiveness of the algorithm.Secondly,select some macro data such as fixed asset investment,rural consumer price index and some financial data such as Vanke's closing price,Wasu's closing price to fit the time series model.At present,the research theory of the fitting method of the model is relatively mature.The ACF,PACF and other indicators are used for preliminary judgment.And the statistical software R is used to identify the model with the smallest AIC.Besides,the LB test and the QQ map are used for the residual test.And McLeod.Li.test,LM test are used to verify the ARCH effect.Then apply the algorithm to identify the change point.Finally,the change point identified by the algorithm are analyzed.Review the changes in national macroeconomic policies and other relevant data,then find out and analyze the causes of the changes.It provides some guidance for the change of time series data,and at the same time,it makes reference for forecasting economic trend and preventing financial risks.This is also the significance of research on change point.