Prediction Of Volatility Of Shanghai And Shenzhen 300 Index Based On Interval Regression Model

Investors are always eager to bear small risks to achieve large benefits in financial market,so a more accurate model for measuring market risks is needed.In most traditional time series analysis,models are built on basis of closing price and theory of random volatility model,from the perspective of probability.The fluctuation of prices of financial assets in stock markets provides information by means of the internal data sequence,[bottom price,ceiling price].Depending only on singular value sequence results in inevitable loss of observed information,and then reduces the effectiveness of models.There are few researches that introduce interval models into financial field.Therefore,the dissertation constructs interval models,based on financial interval data to forecast stock price and volatility.Volatility is characterized by fat tail,leverage,asymmetry,cluster,etc.Therefore,for better interpretation of leverage effects of volatility,many researches have applied smooth transition regression model to volatility model of single value sequence(like HAR model).The research of model integrated with smooth transition regression boasts preferably explanatory power and fitting.There still exists a blank area in the research of interval model reasonably combined with smooth transition.The dissertation builds interval autoregressive model,CM,CRM,CCRM,PM,for financial interval data,and introduces smooth transition function into regression model within interval radius.Smooth transition function enables the author to consider 3 different state variables and build 12 models.In the case demonstration part,it selects daily interval price series from Shanghai and Shenzhen 300 Index of stocks between January 2,20018 and December 29,2017 as its data for fitting and assessment of models.According to stock market volatility,the author classifies all samples into normal fluctuation and exceptional fluctuation.Lastly,the author makes a comparison of the predictive abilities of CM model,CRM model and PM model.Besides,CM model and CRM model are exceptions of PM model,and further illustrates the improvement of CM model by CRM and PM using DM test.Then,this dissertation utilizes DM testing and MCS testing,compares out-of-sample predictive ability of interval model and GARCH model via various loss functions,and evaluates fitting and predictive ability of volatility between interval models.Overall,in the stage of normal volatility,the derivative model of CRM model with the application of smooth transition,CRM-ST(d)model,markedly improves predictive ability of CRM model,and meanwhile,the derivative model of PM model with the application of smooth transition,PM-ST(d)model,markedly increases predictive ability of PM model.However,in the abnormal fluctuation phase,the CRM-ST(w),CRM-ST(m),PM-ST(w),and PM-ST(m)models show better predictability of volatility.