Parameters Estimation Of Change Point Problem Of AR(P) And ARMA(P,Q) Basaed On Bayes Method

Mainly study on parameters estimation of change-point problem of AR(P) and ARMA(P,Q) model basaed on Bayes Method. The introduction part introduce the development situation of change-point problem domestically and abroad, and make a definition for the change-point problem in time series; In chapter 2, we get all parameters of estimation expressions of AR(P) by using Bayes method, including the position of change point k, start from AR(2),then gen-eralize to AR(P)(AR(1) has been discussed by LiuQin). In chapter 3, we get k and other parameters of estimation expressions of ARMA(P,Q) by using Bayes method; In chapter 4, we discussed the MLE method for estimating all parameters in AR(P) and ARMA(P,Q) change-point model, and introduce the SIC method for confirmation of change point. In chapter 2 and 3, we provide numerical simulation and compare Bayes method with MLE method, and get good result. In the forecast research, we found that the forecast results will be better when we consider the change point in series than we don't consider the change point in series, and we also find using Bayes method can get better forecast results than MLE method.