NN-elman-GARCH Family With The Case Study Of China Stock Market

Arrangements in the portfolio, as well as the financial risk management in theoryand practice, we regularly come into contact with asset volatility as a measure,representing the risk profile of the assets. In modern financial theory, there are twobasic activities, namely, risk assessment and asset pricing, which are inseparable fromthe measurement of assets income rate volatility, so that accurate measurements, as wellas forecasting volatility of financial assets is one of the important issues in the study ofthe financial markets.For years, scholars at home and abroad on asset volatility measurements conducteda lot of research, and significant research achievements have been made. Actually, sinceEngle1982first made autoregressive conditional heteroscedastic variance theory (thatARCH model), this model has been widespread, In order to portray financial assetsproceeds rate of observation characteristics-including: fat-tailed, clustering, and etc,researchers began to improve the ARCH model, such as GARCH, EGARCH, TGARCH,GJRGARCH, NAGARCH, APARCH model and etc. And the most famous one isGARCH model made by Bollerslev in1986, namely generalized Autoregressiveconditional heteroskedasticity model. But as we know, parametric GARCH model hasits biggest drawbacks, which requires specific assumptions and conditions a parameter,thereby causing the limitations of models. Researchers in order to break theseassumptions limit and presented another new method, namely nonparametric GARCHmodel. However, researchers found that the nonparametric method also has numerousflaws of its own, such as excessive dependence on large samples of data, dimensionsdisaster and lack of explanation. So in the last few years of study, many researcherstried using Semiparametric GARCH model to measure the volatility of financial assets,which not only has better fitting and forecasting capabilities, but also has theexplanation from parameter method.Semiparametric GARCH models are usually consists of GARCH model and hybridinto another non-parametric methods, for example: Wavelet analysis, local polynomialregression, Copula functions, and so on, and this study on volatility of Chinese stockmarket based on artificial neural network algorithms and Semiparametric GARCHmodel family of mixed methods. In this paper, we uses a new combination withGARCH models and artificial neural network algorithms to summarize and compare the Semiparametric ANN-GARCH models family and traditional parameters GARCHmodels family ethnic differences in analysis and forecasting capabilities in the Chinesemarket, and get better results.