Research On The Volatility Of High-frequency Data Based On The HAR Fam-ily Model

With the continuous innovation and development of the financial market,many factors superimposed on the market trend,and market volatility has become frequent and intense.How to grasp the volatility changes has become an urgent problem to be solved,and the volatility is one of the important indicators in the study of financial markets.One.The research on volatility first started in the field of low-frequency data,and often expanded research based on the simple and easy-to-operate GARCH model.As high-frequency financial data becomes more accessible,how to measure and measure in the field of high-frequency data Forecasting volatility has also become the focus of research,and most of them will be improved based on the HAR-RV model constructed based on the theory of market heterogeneity.Based on this,this article gradually explores the various properties of volatility and improves the prediction model from data sampling frequency,modeling methods,and influencing factors.This article first compares the prediction effects of low-frequency data modeling and high-frequency data modeling,and the results show that the performance of the model built with high-frequency data is better.Then the two widely used models,GARCH model and HAR-RV model,are compared for their ability to predict volatility when using high-frequency data for modeling.The results show that the HAR-RV model performs better.Therefore,based on the HAR-RV model,this article further considers the impact of measurement errors and market microstructure on the volatility forecast.The above two factors are introduced into the model to improve,and the HARQ model and the HARQ-N family model are expanded..In addition,considering that there will be different measurement forms for the realized volatility,this paper also compares the estimator TSRV and the estimator MPRV,and constructs HARQ-N?1 and HARQ-N?2 models respectively.Both the stochastic simulation and empirical analysis results show that the model improvement that introduces measurement errors and market microstructure noise is effective,can improve the prediction accuracy of the model,and the estimator TSRV is better than the estimator MPRV.The HARQ-N?1 model constructed using TSRV estimates has the best effect in predicting market volatility and can help us more accurately grasp market volatility.