Repository files navigation

fable.tbats is a wrapper around the implementation oftbatsoriginally from theforecastpackage, though it now works without this dependency.

pak::pkg_install("JSzitas/fable.tbats")

Used just like any model infable:

library(tsibbledata)library(fable)library(fable.tbats)library(dplyr)# fit models to the pelt dataset until 1930:train<-pelt %>%   filter(Year<1930)test<-pelt %>%   filter(Year>=1930)models<-train %>%   model(ets= ETS(Lynx),bats= BATS(Lynx),tbats= TBATS(Lynx)         )# generate forecasts on the test setforecasts<- forecast(models,test)# visualizeautoplot(forecasts,pelt)

Similarly, accuracy calculation works:

train_accuracies<- accuracy(models)knitr::kable(train_accuracies)

test_accuracies<- accuracy(forecasts,test)knitr::kable(test_accuracies)

As does refitting:

models<- refit(models,pelt )

Of the functionality available in theforecast package, onlyforecast::tbats.components() is missing.

Fitting bats/tbats to a few long series can (potentially) be slowerusing this wrapper than using the forecast package. This is due to thefact that the internal tbats/bats algorithm always executes sequentially(i.e. withuse.parallel = FALSE ) to prevent issues with nestedparallelism (as thefabletools::model function is taken to beresponsible for handling parallelization).

This should never be a problem on many time series, but does lead to asignificant slow-down if you are only modelling a single/few timeseries. Nonetheless, in those cases thefasster package might be muchbetter suited for your use case anyways.