Research On Efficiency Evaluation Of GARCH Prediction Model Based On Different Distributions

Under the influence of economic globalization,more and more people participate in the financial market.Because financial market participants often focus on the future development trend of financial index,time series prediction model has become the first choice for decision-makers because it can fit the prediction model according to the characteristics of past data.However,in the financial market,volatility is often the primary consideration of decision-makers in the game with the market.Therefore,people pay more attention to how the value fluctuation of the securities market changes.By grasping the value fluctuation characteristics of financial assets,we can reduce the operation risk and maximize the profit.In practical application: Modeling and data analysis of financial market data by using the simulation of volatility can help investors better grasp the asset value in the market,minimize losses,and effectively measure and prevent financial market risks.As a modern financial event sequence model,grach model is based on the characteristic of volatility aggregation.Volatility aggregation is that there is a certain relationship between the current volatility and the past volatility,and the concept of variance is correspondingly extended to conditional variance.The so-called conditional variance refers to the variance with known information of the past time.Although GARCH model solves the problem of inaccurate prediction of variance in financial time series,there is still the problem of inconsistent fitting of fluctuation variance distribution in financial time series.Therefore,many researchers have conducted a series of studies on the fluctuation distribution of financial market data,but different distributions always correspond to the situation of single data.In many cases,decision makers do not know how to choose the distribution form to support their decision-making needs.Starting from the analysis of GARCH models with different distributions,for the purpose of serving the investment needs of decision-makers,and with the help of the mature framework of DEA model,this paper evaluates the efficiency of five common GARCH models,in order to provide reference for decision-makers to choose suitable distribution models to fit financial data.This paper selects five thick tailed distributions commonly used in the study of financial data: t distribution,St distribution,GED distribution and nig distribution,and adds Levy asymmetric distribution to participate in the comparison.At the same time,the three evaluation indexes of goodness of fit,deviation and prediction direction accuracy are considered as the input and output of the efficiency comparison of subsequent DEA models.The super efficiency unexpected DEA model is constructed to obtain the ranking of different distribution models.In addition,in order to obtain a more three-dimensional efficiency ranking and the gap between different distribution models,The background dependent DEA model is also used to calculate the improvement value and attraction value of different models and different production fronts,so as to evaluate the prediction model and obtain a complete efficiency ranking of the prediction model.After a series of efficiency evaluation,it is concluded that among the GARCH models with different distributions,nig distribution and GED distribution are always the prediction models that are most in line with the accuracy evaluation index,that is,when the nig and GED distribution prediction models have large volatility in the stock price,they are still better than other distribution prediction models in predicting the future fluctuation direction.After DEA efficiency evaluation model,different GARCH prediction models can be evaluated according to the needs of decision makers,and more consistent ranking results can be obtained under different evaluation indexes.