Closed Dooruj closed 6 years ago
@Dooruj Sorry about the delay getting to this.
The "forecast" package has a special function meanf()
for producing mean forecasts that you can use in this scenario. This should work:
cv1 <- cvts(AirPassengers, FCFUN = meanf, rolling = TRUE, windowSize = 12, horizon = 2)
extractForecasts(cv1, horizon=2)
This mean forecast example is a special case where a workaround exists, so I'll need to investigate your example to see what is going on here. In general the models/forecasts you make with custom functions need to be very careful to preserve the tsp
timeseries attributes, so that is my guess about what is going on.
The documentation examples have been made somewhat more clear that the tsp
need to be correct (specifically the frequency). Look at the updated example for using "GMDH" for a more useful example instead of the naive forecast toy example here.
This package is very useful. I was running through some of the examples in the documentation, and I might have spotted a bug (But it could be that I am not running the procedures properly, for which I apologise for taking your time). Here's something I tried to run:
v <- cvts(AirPassengers, FUN = "ets", FCFUN = "forecast", rolling = TRUE, windowSize = 12, horizon = 2)
ExtractForecasts works well in this case.
Then I've tried to simply forecast with a constant mean with rolling window:
cv <- cvts(AirPassengers, FUN = "mean", FCFUN = "forecast", rolling = TRUE, windowSize = 12, horizon = 2)
But when I run "extractForecasts(cv, horizon=2)", I get the following error
Error in .cbind.ts(list(...), .makeNamesTs(...), dframe = dframe, union = TRUE) : not all series have the same frequency
Not sure why this should appear. Thanking you,