“预测”包版本3.22 Auto.arima将在R中进行预测，且超过一个季节性周期

Question

I am using ARIMA with a seasonal period a "week" that has 672 measurements like follows:

day_freq<-96
Week_freq<-day_freq*7 #672
## Weekly seasonality
NEW.xx<-ts(x,start=1,freq=672)
F.Xreg <- fourier(NEW.xx,24)
fit <-auto.arima(NEW.xx, D=0, max.P=0, max.Q=0, xreg=F.Xreg)
##new xreg fourierf
FForecast<-fourierf(NEW.xx,24,3000)
forecast(fit, h=3000, xreg=FForecast, level=0)

How can I combine this two seasonalities and use a dayly seasonality "96" in the same time as a Week seasonality together?

Answer 1

First, ARIMA models do not handle large seasonal periods very well. Second, R does not allow for multiple seasonal periods in an ARIMA model. I suggest you use the tbats() function in the forecast package which handles long seasonal periods and multiple seasonality.