enhancementgood first issuelib:fracdiffmodule:forecasting
Repository-Metriken
- Stars
- (7.162 Stars)
- PR-Merge-Metriken
- (Durchschn. Merge 26T 10h) (86 gemergte PRs in 30 T)
Beschreibung
Is your feature request related to a problem? Please describe.
ARFIMA and various models for fractionally differenced models are useful in modelling low signal-to-noise ratio time series. Implementing fractional differentiation would be a useful addition to sktime.
Describe the solution you'd like
Implemented fractional differencing, either as an expansion to the Differencer class, or as a separate class. This basically reduces to copying the deprecated fracdiff library code. It has scikit-learn compatible transformer, so adding it to sktime should be quite easy.
Describe alternatives you've considered
I used to use fracdiff library, but now it's unmaintained.