The Prediction Of The Daily High And Low Prices Based On Fractional Cointegration

Daily high and low stock market prices provide valuable information about range-based volatility which can not be reflected by the closing price, and can be as the reference value of the stock market future development forecast values, its importance can not be ignored. From the domestic view of the prediction research of daily high and low stock market prices, mostly based on the two existing between integer dimension cointegration relationship, but a large number of studies have been indicating the presence of significant long memory characteristics of Chinese stock market. At this time, traditional integer dimension cointegration will be difficult to characterize the sequence between the long-run equilibrium relationship, academic admiral cointegration analysis framework is extended to the fractal dimension should be more reasonable than the integer dimension.In this paper,we choose the daily high and low prices of Hushen 300 index? Shanghai stock index and Shenzhen stock index in Chinese stock market as the research object. Firstly, using the KPSS test, the long memory of the semi parametric ELW estimation methods of two price time series of test and parameter estimation. Then, the use of N-S and J-N these two kinds of situations for fractional cointegration cointegration rank test method on the sequence of the existence of fractional cointegration test. Finally, using FCVAR model based on fractional cointegration and cointegration based on integer dimension VECM were fitted to estimate the two price series, make relevant based on the error prediction, and the prediction accuracy of the two models of the evaluation of the results.The empirical results show that:the daily high and low price and its price range sequence of Hushen 300 index?Shanghai stock index and Shenzhen stock index in Chinese stock market are long memory; the daily high and low sequence of these three index are fractional cointegration; the daily high and low price can be modeling of estimation and prediction. Based on the relative error of prediction, results of FCVAR model based on fractional cointegration is smaller than VECM integer dimension cointegrated, the accuracy higher.