Mix-Garch-L Model Based On Adaptive Weight Function And Its Application

With the increasing popularity of the application of high-frequency data,the introduction of high-frequency data analysis into volatility research has gradually become one of the current hotspots.The more widely used is the volatility model based on mixing sampling technology(MIDAS)and GARCH model.The Midas method used in this kind of volatility model is only related to the sequence of data,and its weight distribution method is not suitable for refining the effective information of daily high-frequency trading data.The reason is that the Midas method gives the same weight mode in different trading days,and the trading characteristics in the process of high-frequency trading in different trading days are real-time changes.The time of high-frequency trading that has a greater impact on the future volatility is not fixed in the trading day.Forcibly using Midas method,that is,using a fixed mode that only depends on the sequence of time intervals to give corresponding fixed weights to high-frequency transactions in different trading days,will inevitably lead to the mismatch of weights and the distortion of refined information.Therefore,it is best to use a new method to extract the weight of highfrequency trading data automatically and effectively,so it is best to use a new method to extract the weight of high-frequency trading data.By introducing a new kind of adaptive weight function,this paper proposes a new data extraction method suitable for the fluctuation of high-frequency data,extracts the effective information in the current high-frequency data,combines it with the lowfrequency data,and constructs a fluctuation model that can make full use of the highfrequency and low-frequency information: mix-garch model.Because the new weight function is in logarithmic form,the model is recorded as mix-garch-l model.Because the newly proposed mix-garch-l model is derived and improved on the basis of the original GARCH model and changes the expression form of the original GARCH model,it is necessary to estimate the parameters of the new model.In this paper,the parameter estimation method of mix-garch-l model is given,the theoretical properties of the estimator are analyzed,the corresponding central limit theorem is proved,and the data performance of the estimator is simulated and tested by service boost method.The new volatility model has the following advantages:(1)The weight is given by the weight function according to the different order of the traditional data,and the weight is different from the traditional data.The independent variable of the new weight function is the variable that can describe the characteristics of high-frequency trading,which makes the data fusion method based on the new weight function better allocate different weights according to the trading characteristics.This allocation method can better automatically adjust the weight assignment in different trading days according to the changes of trading characteristics,Thus,the weight assigned to each high-frequency trading day is consistent with the impact effect of future volatility.(2)The new volatility model can make full use of a variety of high-frequency trading data in the same trading process,which will make mix-garch-l model have better prediction accuracy and prediction advantages.The practical application results in this paper also prove this point.When analyzing the return rate of Shanghai Composite Index,a variety of models are used to predict the volatility of actual data.The results show that mix-garch-l model can predict the volatility more accurately and steadily.The proposal of mix-garch-l model also expands the method of using mixing data to analyze volatility and other related problems.