Prediction Research For Chinese Stock Market Volatility Of High Frequency Data

This paper mainly studies the prediction research for the Chinese stock market volatility of high frequency financial data.First of all, this paper introduces the high frequency financial data, it is compared with the previous year, month, week, etc for the sampling interval of low frequency in terms of financial data, which usually take days, hours, minutes and even seconds as the frequency of collected according to the time order of financial data sequence. In this paper, we study the data of the sample using the benchmark Shanghai composite index on January4,2010to June29,2012the high-frequency data of every1minute.Secondly, the paper uses realized volatility to estimate the actual volatility, it theoretically deduces the calculation method of realized volatility and realized volatility are presented in the limit properties, it is an important theoretical base for study in this paper. Next, this article refers to the time sequence of an important properties of long memory, gives the long memory three definitions from the perspective of temporal and spectral density and introduces the long memory of a test method-R/S analysis method. Then, this paper mentions the ARFIMA (p, d, q) model, put forward the significance of the fractional order differential, and through the use of Stirling formula gives a simple method for fractional order differential, through simple R language program can be realized.Finally, an empirical study of the high frequency financial data, for logarithmic realized volatility established ARFIMA (1,0.22,2) model, and carries on the10steps backward prediction, we know that the model is appropriate by predicting the results. And by comparing the modeling results of the fractional order differential and1order difference, we can know that the fractional order differential method which put forward by this article is effective and meaningful.