This message was deleted.
# neural-forecasts
Slackbot
12/27/2023, 8:38 AMThis message was deleted.
c
Cristian (Nixtla)
12/27/2023, 4:49 PMHi! You should remove utc
m
Marco Zucchini
12/27/2023, 5:13 PMHi @Cristian (Nixtla) sorry to bother you: as I have removed utc ( as written before I had already tried) but it leads to ValueError (only) when I do nf.predict_insample : _You are trying to merge on datetime64[s] and datetime64[ns, America/New_York] columns. If you wish to proceed you should use pd.concat --> the date I am using comes from yfinance data and converted as above with pd.todatetime (pandas 2.0.3) and it is like that_ 2018-12-27 000000-05:00 --> even if utc=False it indicated _tz:
DstTzInfo 'America/New_York' LMT-1 day, 19:04:00 STD --> pandas reference 2.1.4 give perhaps some hints about utc https://pandas.pydata.org/docs/reference/api/pandas.to_datetime.html --> is it a conversion problem? why it works with nf.predict and not with nf.predict_insample? thank you!
c
Cristian (Nixtla)
12/28/2023, 4:49 PMIt seems that your data still has the timezone (it says America/New_York] in the error. Can you try with ``df['Date'].dt.tz_localize(None)`` . In this post they cover more options: https://stackoverflow.com/questions/49198068/how-to-remove-timezone-from-a-timestamp-column-in-pandas
m
Marco Zucchini
12/29/2023, 3:45 PMHi @Cristian (Nixtla): it sounds you have found the workaround. Thank you!
🙂 1