Research On Modeling Intraday Volatility Of High-frequency Financial Time Series

Volatility is one of the research hotspots of financial economics,which is widely used in the fields of financial asset portfolio selection and risk management.In the early of volatility modeling which was based on low-frequency data,the traditional GARCH and SV models were mainly adopted,which can well depict and describe the volatility of low-frequency data.For the modeling of high-frequency data,the traditional GARCH and SV models can’t achieve good results.Therefore,realized measure(RM)represented by realized volatility(RV),has been widely applied to the fluctuation estimation of high-frequency data due to its model-free,simple calculation and other characteristics.On the basis of realized volatility,HAR models are established to describe and predict future volatility.Although these models contain more information based on high-frequency data,the final result also is daily volatility.Obviously,such daily volatility is unable to describe the intraday level of the volatility model.This paper mainly studies intraday volatility of high-frequency data from the perspective of market microstructure noise.Firstly,some characteristics of high frequency data which are different from low frequency data are introduced and briefly analyzed.Then,the traditional GARCH model is recommended.When the GARCH model is used to depict the intraday volatility of high-frequency data,the GARCH model does not take into account the market microstructure noise characteristics of high-frequency data,which is one of the objective and non-negligible characteristics.Market microstructure noise interferes with the return rate sequence by affecting the logarithmic price.The influence of market microstructure noise can’t be ignored when the intraday volatility of high-frequency data is studied.Therefore,on the basis of GARCH model,this paper introduces market microstructure noise as an important variable into the mean value equation of GARCH model,and establishes the GARCH model with market microstructure noise.In the parameter estimation of the model,the maximum likelihood estimation method is used to solve the parameters in this paper.Finally,combined with the prediction methods and evaluation criteria of intraday volatility,the validity and accuracy of the model in this paper are verified by simulation experiments and empirical studies.Monte carlo simulation is used to test the effectiveness of the GARCH model and GARCH model with marketmicrostructure noise variable.The experimental results show that compared with the GARCH model,the proposed model is more accurate in solving model parameters,and the estimated parameters are closer to the set real values in this paper.Then,the out-of-sample prediction and in-sample fitting results are compared with the traditional GARCH model based on the empirical research and analysis of the5-minute high-frequency data of the CSI 300 index and the NASDAQ composite index.The empirical results show that the GARCH model with market microstructure noise can better describe the intraday volatility of high-frequency data.The experimental and empirical results show that the proposed model has better and more accurate parameter estimation and prediction ability.