Comparative Analysis Of Risk Prediction Models For China's Crude Oil Futures Market

Crude oil is an significant guide for the commodity market in various countries,at the same time it is also the core part of industrial development.The Asia-Pacific region,headed by China,has developed rapidly recently,and the demand for crude oil is constantly speeding up day by day.In 2018,China's crude oil futures were officially appeared on the market,providing crude oil stakeholders with a new investment and hedging tool,and a new way for institutional asset structure allocation.So how to make good use of crude oil futures tools has become an issue of concern to investors.The study of risk prediction on the crude oil futures market plays a vital significance,which on theoretical and practical,for my country's crude oil futures investors,arbitrageurs and oil policy makers.Since Shanghai crude oil futures is an emerging market,most of the domestic research on this market at this stage is qualitative research,and there is very little research on the risk of my country's crude oil futures market.Therefore,this article fills in the gaps in the market risk measurement research A variety of models are used to measure market risk,and the results are compared and analyzed in stages.On this basis,a combination model is also tried to predict risk.Specifically,in order to find a risk measurement model that is consistent with China's crude oil futures market,this article uses the logarithmic return data of the scm.INE closing price as the research object,adopts a variety of current mainstream Va R forecasting methods for empirical analysis,and compare the accuracy of model estimates by backtesting.The models we uses in text include several historical simulation methods,various GARCH models,extreme value theory,and quantile regression models of adding different explanatory variables.Considering the differences in the estimation effects of different models in different market stages,this article further divides the forecasting stage into two stages: market turbulence and market plateau.Therefore,we will analyze the measurement effects of different models in different stages.Due to the differences in the models suitable for different market stages,based on a single forecasting model we tried a combination model for risk prediction.Through empirical research,it is found that: 1.Comparing the test results of various single models during the entire forecast period,the forecast effects of different models are different.During the entire forecast period,the HS-TW model has the lowest failure rate at the 95% confidence level,and the GARCH model The best performance at the 99% confidence level;2.From the estimation effect of different market stage,when the market environment is in the extreme change stage,at the 95% confidence level,only QGARCH,GARCH-POT models can pass the backtesting,and the failure rate of QGARCH is more closer to 5%;however,at the99% confidence level,only the simple historical simulation method and the volatility-weighted historical simulation method failed the backtesting,which shows that when the market does not show large fluctuations,most risk measurement methods are effective enough to describe the risk changes of our country's crude oil futures market;3.From the point of the prediction effect of a single model and a combination model,the combined model has a significant improvement in the prediction effect of the single model.We can find that the GARCH-POT and QGARCH models all have passed the backtesting under different confidence levels in different market stages.