一、GARCH模型
ARCH模型的建模过程也适用于GARCH模型的建模。在大多数的应用中，只用到低阶的GARCH模型，如GARCH(1,1)模型、GARCH(1,2)模型和GARCH(2,1)模型，因此本文只对比这三种阶数的模型。
二、IGARCH模型
三、GARCH-M模型
四、EGARCH模型
五、模型验证以及预测
library(fGarch)library(rugarch)m<-garchFit(~1+garch(10,0),data=dhp,trace=F)summary(m)resi<-residuals(m,standardize=T)m1<-garchFit(~1+garch(16,0),data=dhp,trace=F)summary(m1)resi1<-residuals(m1,standardize=T)tgarch11.spec = ugarchspec(variance.model = list(model="fGARCH",submodel="EGARCH", garchOrder=c(1,1)),mean.model = list(armaOrder=c(0,0)))#EGARCH fit <- ugarchfit(spec=tgarch11.spec, data=dhp,solver='solnp') fitspec1=ugarchspec(variance.model=list(model="iGARCH",garchOrder=c(1,1)),mean.model=list(armaOrder=c(0,0)))#IGARCHmm=ugarchfit(data=dhp,spec=spec1)fore = ugarchforecast(mm,50)source("garchM.R")mm<-garchM(dhp)#GARCH-Msummary(mm)resi<-residuals(m,standardize=T)#残差走势图time<-c(1:3000)/240+plot(time,resi,xlab="year",ylab="stand-resi,",type="l",lty=1,col="blue",main="标准化残差时序图")acf(resi,lag=40)pacf(resi^2,lag=40)a<-predict(m,100)#预测names(a)a1<-a$meanError#预测波动率走势图date<-c(1:100)/240+.3583plot(date,a1,xlab="time",ylab="预测的波动率",type="l",lty=3,col="blue",main="预测的波动率走势图")pred.model <- predict(m, n.ahead = 10, trace =FALSE, mse = 'cond', plot=FALSE)View(pred.model)
