Influence Analysis Of Change-point Estimation On Prediction Of State Space Model

In this dissertation, the influence analysis of change-point estimation onprediction of state space model is mainly studied.In the first chapter, the research background and status at home and abroad ofchange-point and state space model are described. Several classical detectionmethods of change-point are given. In the second chapter, state space model andKalman filter are introduced. And the standard form of state space model is studied.In the third chapter, the detection method of-distribution with change-point isdiscussed. The influence on prediction of state space model of-distribution withchange-point is also analyzed.First, the all rise and fall rate of return in the closing price of ShanghaiA-stock index are given. Furthermore, we analyze the all rise and fall rate of returnby the theory of Γ-distribution with change-point to get the number and thelocation of change-point. Then, according to the number of change-point, statespace models for three different stages of the closing price of the Shanghai A-stockindex are established. It is concluded that the fewer number of change-point, thebetter the state space model has the accuracy of prediction by comparing theprediction results. Finally, it shows that the prediction effect of state space model isbetter than ARIMA model and autoregressive model when there is no change-point.