While checking the code for splitting the data into train/val datasets, I came across this code.
I think this logic of splitting works well for wide format but not long format. For long format we should group by the series_id column, do the filtering and then concatenate the data again.
While checking the code for splitting the data into train/val datasets, I came across this code.
I think this logic of splitting works well for wide format but not long format. For long format we should group by the
series_id
column, do the filtering and then concatenate the data again.Can you please look into this?
Thanks