Prediction And Test Of High Frequency Fluctuation Rate Of HS300 Index Based On HAR Family Model

The stock market is a barometer of China's economic development and the most important capital operation for investors.It not only determines the stability of the economy but also the result of the investment and financing of enterprises.It has always been the core area of the financial system.With the imperfect investment channels in our country,the a reasonable operation of the stock market plays a crucial role in maintaining order in the financial market and stimulating economic development.Volatility is an important parameter used to describe the uncertainty of financial assets,which reflects the stock market's operating conditions.Accurately forecasting the volatility can avoid the losses to investors due to systemic risk and non-systematic risk.For stabilizing the stock market,Provide preconditions for the implementation of government control measures.Therefore,choosing an accurate volatility model has become a necessary condition for research.Based on this,this dissertation starts from the current research of volatility at home and abroad,combs the related theories of volatility,and analyzes the most suitable HAR family model theory and its test model SPA based on the widely used high frequency data research methods Testing method.Taking into account the stock market benchmark in Shanghai and Shenzhen 300,this paper collects high frequency data of the CSI 300 Index for 5 minutes.Based on the HAR family model,which has realized the volatility modeling at high frequency,the paper uses rolling time window The out-of-sample prediction and the Bootstrap-based SPA test method,respectively,using six commonly used loss function as the evaluation criteria,a comprehensive comparison of nine common volatility models for the Shanghai and Shenzhen 300 Index day,week and month has realized volatility forecast accuracy.The results show that the stock market in our country has long memory,and the HAR-RV-CJ model has the highest accuracy for out-of-sample prediction of short-term and medium-term volatility.The logarithmic form of the HAR-RV-CJ model forecast has the highest level of accuracy and better explain future long-term volatility.