Volatility Prediction Of Gem Index Based On High Frequency Data

It is difficult for people to fully understand the characteristics of the volatility of financial market because of the unpredictable financial market,so it is very accurate and necessary to reflect the characteristics of market volatility by analyzing high-frequency financial data.With the update and iteration of computing data processing tools,and the continuous development of corresponding computing data processing methods,the cost of data recording and storage has been significantly reduced,and the international financial market has also emerged the analysis and Research on the current high-frequency financial data fluctuations.High frequency financial data refers to the trading fluctuation data of the financial market in the day.It refers to the financial trading fluctuation data calculated by high frequency sampling during the opening period.The research of high frequency financial data can reflect and calculate the current fluctuations of various financial products in the international financial market more carefully and comprehensively.Through the analysis and research of high frequency financial data,It can better help people to avoid the impact and risk of the current financial product fluctuations on the investors’ earnings,and help investors to quickly grasp the characteristics of the current financial market fluctuations and the internal principles.This paper is mainly from the perspective of high-frequency data to study the change rule of gem index,through the five minute data for its volatility analysis and empirical research,in order to improve the stock index and measurement,this paper constructs more effective characteristics and portfolio model two directions.The first is to build the indicators required by the realized model.After analyzing the characteristics of high-frequency data,the corresponding daily realized volatility,weekly realized volatility and monthly realized volatility data are constructed.The second is to build the corresponding indicators of SVM,including basic indicators and technical indicators,and to normalize and standardize the data to make the data more scientific.The last is to build the technology The index makes SVM model,svm-garch model predicts closing price,calculates volatility according to closing price,to verify the fitting degree and prediction ability of data outside and inside samples to each model.Through comparative analysis,it is found that svm-garch model is the most correct one.The results show that: first,for gem index,the prediction effect of traditional measurement model and machine learning method is different,and the effect of machine learning method is significantly stronger than that of traditional measurement method;second,for nonparametric model of high-frequency volatility,the estimation is more accurate after introducing jumping elements,that is,HAR-RV-J has higher prediction accuracy than HAR-RV model,and The external weighted realized volatility model(HAR-RRBV)is more accurate than the conventional realized volatility model(HAR-RV);thirdly,the combined model SVM-GARCH model performs well in the prediction ability,and its prediction accuracy is better than the simple SVM model.