Time series coming from Gapfilling vs. features_from_raw_dates learning samples does not match
Hi all,
I did a quick test to compare learning samples extracted using 1) Gapfilling mode (default options using iota2/develop) and features_from_raw_dates mode.
After looking at the two time series, the gapfilled time series is surprising in September. I am missing something ? If I understand well, the default gapfilling method is a linear interpolation and should not extrapolate the TS like this.
Thanks, Florian