Open mitchelloharawild opened 6 years ago
Will you consider adding the TSMARS
(Time Series Multivariate Adaptive Regression Splines) model into fable
?
Probably not into fable
, but it certainly would be nice in an extension package.
Hi, thanks for the checklist. I was wondering why are BATS/TBATS not implemented?
I've striked BATS/TBATS out as they are not a priority for the package. It's not to say that they won't ever be included, but they won't included any time soon. This is because we now recommend using the FASSTER model for forecasting time series with multiple seasonality.
Thank you so much! I didn't find the "stlm" model checked here in the fable cran r documentation. Do you have suggestions about where to find it in other sources?
Good question. Decomposition modelling is now generalised for any additive decomposition and can be specified using decomposition_model()
.
There are some examples of STL modelling contained in the docs: https://fabletools.tidyverts.org/reference/decomposition_model.html
Any plan to include X-13-Arima-SEATS?
Sure is. This model was included in the feasts package earlier (as SEATS()
), however we've not exported it for now until we can verify that it is described adequately. Related issue: https://github.com/tidyverts/feasts/issues/66
Hi, I just updated fable to v. 0.2.0 and noticed that the STL function is gone. Will it be added again or was it moved?
STL is in the feasts package, not the fable package.
My bad. Thank you!!
Thank you for this great job. will you consider adding DeepAR, wavenet, nbeats and other probabilistic models from GluonTS toolkit ? (https://gluon-ts.mxnet.io/index.html)
I met with the GluonTS team mid last year, and I would like to have fable wrappers for these models. These won't be included in the fable
package, but possibly in an extension package like fable.gluonts
(similar to fable.prophet
).
I'm currently doing a project where an ARFIMA()
function would be very useful.
I'm currently doing a project where an
ARFIMA()
function would be very useful.
Any generalisations that should be made from forecast::arfima()
?
Actually, I'm thinking it would be better to use the implementation in the arfima package as it is more general and allows for regression terms and transfer function terms.
I have been a huge beneficiary of your work on the forecast package over the years. So first off - thank you so much for all the work you do around this! I saw spline of your list of models and I am anxiously awaiting for its turn to be implemented. So much so that I had hacked the forecast version into a barebones fable version for my own use. Just wanting to show my appreciation and let you know there is at least 1 person out there who would like to see spline implemented.
https://github.com/rando-brando/fable-spline/blob/main/spline.R
Models
Arima
ETS
naive / rwf / snaive
tslm
BATS/TBATSFunctionalReimplementMethodsbaggedModel (https://github.com/tidyverts/fabletools/issues/217)
dshwFunctionalReimplementMethodscroston
stlm
meanf
thetaf
modelAR
nnetar
arfima
splinef
Wishlist
VAR
VARMA