Analysis Of Measurement For VaR Based On Change-point Detection

There are many uncertainties in financial markets,and volatility is an important feature of financial time series.Except to return,investors put more and more attentions on the risk of investment.In recent years,as global financial market has fluctuated more severely,financial risk measurement and risk con-trol become a concern of all investors,including individual institutions,even countries.There are many way to measure financial risk.One common method is the VaR measurement which is proposed by J.P.Morgan.VaR,which means "Value at risk",is the biggest loss an investor will suffer in one period given a confidence level.The calculation of VaR bases on historical data,so it’s very important to choose a suitable period and then use the data in it.We introduce the change-point of financial time series to segment the his-torical data,and just use the data which later than the nearest change-point to measure VaR.The change-point in the paper is mode-change-point,which means the mode of time series’fluctuation has changed before and after the change-point.This paper uses a method,which termed BASTA,to detect the change-point.BASTA, which means binary segmentation for transformed auto-regressive conditional heteroscedasticity,including two stagesrprocess transfor-mation and binary segmentation.The process transformation lightens its tails and decorrelates the original process.And we use the binary segmentation to estimate the change-point.Finally,we present a empirical analysis based on 2031 close prices of Shang-hai Stock Exchange Index.