Parameter Estimation Of Regime Switching Model Based On High Frequency Data

Financial high frequency data is usually divided by a certain time interval,such as hours,minutes,seconds,or even smaller time units,theoretically closer to the continuous-time model.Thus,it is closer to the actual model of financial asset prices than low-frequency data.With the development of technology,the acquisition of financial high-frequency data has become more convenient,and the study of high frequency data has gradually become a hotspot.At present,regime switching of the financial market is usually ignored in the study of high-frequency data and therefore combining the model of high-frequency data with the Markov regime-switching model will be a research direction.The Markov regime-switching model is used to study financial high-frequency data in this paper.Firstly,the Markov regime-switching autoregressive model is combined with financial high-frequency data to model and predict the price by using the filtering algorithm.The closing price of the Shanghai Composite Index during the period from January 3,2018 to March 28,2019 is selected and Markov regime-switching autoregressive model based on the daily closing price(low frequency data)is compared with the model based on the intraday 5-minute closing price(high frequency data).The comparative study finds that the intraday 5-minute high-frequency data has higher accuracy than low-frequency data and can better adapt to the model.Secondly,based on the volatility proxy model proposed by Visser,the Markov regime-switching model is introduced to model and predict the volatility by selecting the realized volatility as the volatility proxy.The regime-switching volatility proxy model is compared with single regime volatility proxy model,and the result shows that the former performs better than the latter in fitting effect.