Research On Financial Risk Prediction In Hong Kong Stock Market Based On DD-EWMA Model

3 Financial risk forecasting has always been a hot issue in complex nonlinear financial markets.It is about volatility,value-at-risk(VaR),and expected loss(ES).Since financial data in different fields contains different information characteristics,it is necessary to use different models to analyze and predict data in different financial fields.Therefore,the study of financial risks in different financial fields has great research value and guiding significance.This paper uses the traditional GARCH model,the data-driven exponentially weighted moving average(DD-EWMA)model and data-driven neural network volatility.These three models are used to forecast and compare financial risks in the Hong Kong stock market.This article selects the stock data under the Hong Kong stock market from January 1,2015 to June 30,2021,which are TCL,Kingsoft,ZTE,Ali Health,Gome Retail and China Railway.First,make point predictions on its volatility;second,predict value at risk and expected loss based on the relationship between volatility and VaR and ES;then,for VaR and ES Make interval predictions and compare.In addition,we study the rolling DD-EWMA fuzzy volatility model,the rolling neural network fuzzy volatility model,and the rolling GARCH model to predict VaR and ES.Finally,the rolling predictions of VaR and ES are back-tested separately to compare the pros and cons of these three models.For back-testing of VaR,we use UC test,IN test and CC test;for ES backtesting,we use the indicator NS.This paper proves theoretically that compared with GARCH model,DD-EWMA model predicts the volatility of financial data without bias;DD-EWMA model predicts the volatility of high-tech stock data with stability.It is empirically shown that:(1)the daily forecasts of VaR and ES are time-varying from these three rolling models;(2)in terms of accuracy,DD-EWMA model and neural network volatility model are more accurate than GARCH(1,1)model;(3)from the effect of the model results,backtesting VaR and ES indicators found that DD-EWMA model has the best performance in predicting stocks with kurtosis characteristics in the Hong Kong stock market;(4)from the comparison of model stability,compared with the neural network volatility model,the DD-EWMA model and GARCH(1,1)model interval prediction width is narrower,so the stability is better;(5)in terms of the operational efficiency of the model,the datadriven exponentially weighted moving average(DD-EWMA)model shorter runtime than data-driven neural network volatility models and GARCH(1,1)models.Therefore,on the whole,using these three models to predict the VaR and ES,DD-EWMA volatility model of data with kurtosis characteristics is the best predictor.