Just an observation and a follow-up question
If I do not delete zero sales observations from the train dataset (as the #3 winners did by:
joined = joined[joined.Sales!=0]
), two things happe:
- yl = np.log(y) issues a warning about likely division by zero
- the m.lr_find() dies without any calculations
Follow-up question:
Has anyone tried to add the zero sales rows back in?
Thanks