熊猫数据透视表中的行/列的百分比

Question

Let's say I've dataframe df :

df.head()
                     M1       M2      M3       M4
Timestamp                                                       
2018-09-20 12:59:57  cat 1    obj_1   name_1   1
2018-09-20 12:58:53  cat 1    obj_2   name_2   1
2018-09-20 12:57:44  else 1   obj_3   name_1   1
2018-09-20 12:57:19  cat 11   obj_2   name_1   1
2018-09-20 12:56:17  cat 11   obj_2   name_1   1

With this df I'm preparing a set of pivot tables for each column presenting both percentage (%) of occurrences as well as it's count (N):

df[['M1']].pivot_table(index=df.index.date, aggfunc=(
    ('%', lambda x: len(x) / df['M1'].count()), 
    ('N', 'size')))

When I come across to preparing pivot table on two series I'd like to display the percentage of occurrences of M1 not in the whole dataframe but in relation to M2 categories. So far I've tried to set the denominator to M2 count, but it's the overall count and not the count of M1 within specific M2 categories:

df[['M1', 'M2']].pivot_table(columns='M2', index='M1', aggfunc=(lambda x: len(x) / df['M2'].count()))

Any clues how to get specific percentage of M1 in each M2 category? Expected output:

M2       obj_1    obj_2    obj_3
M1
cat 1    value1   value*   value*
cat 2    value*   value*   value*
...      ...      ...      ...
cat 11   value*   value*   value*
else 1   value*   value*   value*

where value1 is number of occurrences of cat 1 within all occurrences of obj_1 etc.

Answer 1

You can do a groupby to find number of M2 's for each category, and add it as a column to your dataframe as follows

df['count_M2'] = df.groupby('M2')['M1'].transform('count')

Then you run the pivot_table function to get the percentage of M1 's in each M2 group

df.pivot_table(values=['count_M2'], index=['M1'], columns=['M2'], 
               aggfunc=lambda x: len(x) / x.iloc[0])

---

Details

df

                  Time      M1     M2      M3  M4  count_M2
0  2018-09-20 12:59:57   cat 1  obj_1  name_1   1         1
1  2018-09-20 12:58:53   cat 1  obj_2  name_2   1         3
2  2018-09-20 12:57:44  else 1  obj_3  name_1   1         1
3  2018-09-20 12:57:19  cat 11  obj_2  name_1   1         3
4  2018-09-20 12:56:17  cat 11  obj_2  name_1   1         3

df.pivot_table

       count_M2                
M2        obj_1     obj_2 obj_3
M1                             
cat 1       1.0  0.333333   NaN
cat 11      NaN  0.666667   NaN
else 1      NaN       NaN   1.0