I am using FB Prophet to do a forecasting in Jupyter Notebook using Python 3. My model was working fine, until I decided to split my data up into 'train' and 'test'.
Instead of using 'future = m.make_future_dataframe' I am now doing this: 'future = df_test[['ds']]'
When I make this change, the kernel dies after loading for about 1 minute (Jupyter Notebook). With the old 'future', it loads within a couple of seconds.
This is what my df_test['ds'] looks like:
ds
0 2023-09-28
1 2023-09-30
2 2023-10-06
3 2023-10-06
4 2023-10-10
As you may notice, not every day is present as my data only has a row created every time a change happens (to reduce data). Is this what is causing the issue? I removed the 'HH, MM, SS' part to make sure it follows YYYY-MM-DD.
Many thanks!
I tried troubleshooting with various things, such as the date formatting, having a 'larger' amount of dates in test etc. The kernel kept dying.