Pandas: make pivot table with percentage

Question

I have dataframe

ID,url,used_at,active_seconds,domain
61a77f9e5fd52a50c10cd2d4d886ec68,mazdaspb.ru,2015-01,6,mazdaspb.ru
61a77f9e5fd52a50c10cd2d4d886ec68,maz         


        

Answer 1

IIUC you can use parameter margins for sum values in pivot_table and then divide all values last row All by div:

group = pd.pivot_table(df, 
                       index='used_at', 
                       columns='domain', 
                       values='ID', 
                       aggfunc=len, 
                       margins=True)
print (group)
domain   avito.ru  mazdaspb.ru  vw-stat.ru   All
used_at                                         
2015-01       3.0          3.0         5.0  11.0
All           3.0          3.0         5.0  11.0

print (group.iloc[:-1])
domain   avito.ru  mazdaspb.ru  vw-stat.ru   All
used_at                                         
2015-01       3.0          3.0         5.0  11.0

print (group.iloc[-1])
domain
avito.ru        3.0
mazdaspb.ru     3.0
vw-stat.ru      5.0
All            11.0
Name: All, dtype: float64

print (group.iloc[:-1].div(group.iloc[-1], axis=1) * 100)
domain   avito.ru  mazdaspb.ru  vw-stat.ru    All
used_at                                          
2015-01     100.0        100.0       100.0  100.0

Solution with divide by individual count with div and mul:

group = pd.pivot_table(df, 
                       index='used_at',
                       columns='domain', 
                       values='ID', 
                       aggfunc=len)
          .div(len(df.index))
          .mul(100)
print (group)

domain    avito.ru  mazdaspb.ru  vw-stat.ru
used_at                                    
2015-01  27.272727    27.272727   45.454545

Answer 2

Divide the individual count values obtained with the total number of rows of the DF to get it's percentage distribution as shown:

func = lambda x: 100*x.count()/df.shape[0]
pd.pivot_table(df, index='used_at', columns='domain', values='ID', aggfunc=func)

Answer 3

An alternative approach is to use pd.crosstab, which has similar inputs to the pivotable.

This includes a parameter normalize=False (default setting).

You can change this to normalize=True and it provides the percentages of the total.