For a project I am using a sales dataset available here to make an ARIMA model to predict seasonal sales of their top 5 items. I am currently using a group by to build a table to plug into the auto_arima function but I'm having trouble getting it to work. I know what I want the result to look like but I'm having trouble getting there.

## df_comb is a union of the combined datasets on kaggle w/ year month added in as date

total_sales = df_comb.groupby(['Description','YearMonth'])['Quantity'].sum()

This returns

Description                          YearMonth
  DOORMAT UNION JACK GUNS AND ROSES  2010-01      150
                                     2010-02        5
                                     2010-03       14
 3 STRIPEY MICE FELTCRAFT            2009-12      261
                                     2010-01       64
                                                 ... 
ZINC WIRE SWEETHEART LETTER TRAY     2011-08        1
                                     2011-09        2
                                     2011-10        8
                                     2011-11        9
                                     2011-12       13
Name: Quantity, Length: 61768, dtype: int64

I have no idea where to go from here to get the top 5 and then for my ARIMA model do I plot each individually?

Am I thinking about this all wrong?