Change Point Detection Of Time Series Models And Application In Early Warning Monitoring

Time series are common in various fields.Therefore,more and more scholars have begun to pay attention to the research and application of it.The change point is one of the hot topics.In actual production and life,People hope to detect the time when the mutation occurred to make preparations for it and minimize the loss.In this paper,firstly,the ARMA model and the ARCH model are detected by Bayesian method.After the appropriate conjugate prior distributions are selected,according to the Bayesian principle,the posterior distribution of change points is obtained through iteration or sampling method.Moreover the effect of change point detection is measured by the ROC curve.This paper presents the effects of multiple sets of data validation algorithms for the two models.Finally,two cases of application of the algorithm are given.The first one is applied to the change analysis of the closing prices of the two stocks of Ping An Bank and Sugon.The comparison between the real mutation and the actual mutations yields the most suitable model for fitting the time series of stock index and the most suitable detection method.In addition,this paper also analyzes the causes of the mutations in the stock market.The second is to apply this method to the indicator data of Jiangsu's total output and obtain some results that are different from the stock market data.