A Research And Analysis Of High Frequency Data Volatility Based On Chinese Financial Market

The world economy has been growing rapidly,and the financial product prices have become more volatile in financial markets.In order to avoid the risks caused by these fluctuations,a large number of financial scholars and investors analyse and study the volatility of financial markets.Due to the complexity and unpredictability of financial markets,it is difficult for people to understand the law of its change.Therefore,analyzing the characteristics of financial fluctuations is conducive to grasping the essence and internal principles of financial markets.It should be primacy that analysing the characteristics of financial time series,when we analyse the characteristics of financial market fluctuations.the characteristics of financial time series vary due to different financial markets,and the time series is an uncertain sequence.So analyzing its volatility plays a decisive and practical role in promoting the study of the microstructure of the financial market.This thesis mainly analyses and empirically research the volatility of CSI300 stock index futures from the perspective of high-frequency data research.First of all,it is verified that the high-frequency data of CSI300 stock index futures have the typical statistical characteristics of high-frequency data,namely,the heavy tail,calendar,long memory and autocorrelation.Then,an empirical comparative study of the four volatility rates that is realized volatility,realized range-based variation,realized bipower variation and realized range-based bipower variation is conducted in the descriptive statistics,jumping fluctuation characterization and long memory.The results show that,realized range-based bipower variation is better than others.On this foundation,heterogeneous autoregressive model based on realized range-based bipower variation(HAR-RRBV)and a heterogeneous autoregressive model based on realized range-based bipower variation considering jumping(HAR-RRBV-J)are constructed.In the empirical part,the comparison is mainly based on the degree of fit within the sample and the out-of-sample predictive power.The out-of-sample forecasting uses dynamic predictive multi-step static forecasting.Through intuitive graphical trend analysis and root mean square error,average absolute error,average absolute error percentage,and heteroscedasticity adjustment mean square error,the loss function values are compared and analyzed.The prediction effect between the models finds that the HAR-RRBV and HAR-RRBV-J models can effectively improve the accuracy of predicting the volatility of the CSI 300 stock index futures.At 1% confidence level,it is more accurately to predict risk by using realized range-based bipower variation.