Pandas dataframe，如何按多列分组并为特定列应用总和并添加新的计数列？

Question

Given a dataframe df1 as follows:

Col1    Col2    Col3    Col4    Col5
-------------------------------------
A       1       AA      10      Test1
A       1       AA      5       Test2
A       2       AB      30      Test3
B       4       FF      10      Test4
C       1       HH      4       Test7
C       3       GG      6       Test8
C       3       GG      7       Test9
D       1       AA      4       Test5
D       3       FF      6       Test6

• I want to group by Col1, Col2 and Col3 and

• Add new column Count: size of each group

• Add new column Col4_sum: sum of each Col4 in each group

---

Output need

Col1    Col2    Col3    Count   Col4_sum
----------------------------------------
A       1       AA      2       15
A       2       AB      1       30
B       4       FF      1       10
C       1       HH      1       4
C       3       GG      2       13
D       1       AA      1       4
D       3       FF      1       6

I try to use

df1.groupby(['Col1','Col2','Col3']).size 

but get only Count column.

Answer 1

Use GroupBy.agg with tuples for specify aggregate function with new columns names:

df = (df1.groupby(['Col1','Col2','Col3'])['Col4']
         .agg([('Count','size'), ('Col4_sum','sum')])
         .reset_index())
print (df)
  Col1  Col2 Col3  Count  Col4_sum
0    A     1   AA      2        15
1    A     2   AB      1        30
2    B     4   FF      1        10
3    C     1   HH      1         4
4    C     3   GG      2        13
5    D     1   AA      1         4
6    D     3   FF      1         6

In pandas 0.25+ is possible use named aggregation :

df = (df1.groupby(['Col1','Col2','Col3'])
         .agg(Count=('Col5', 'size'), Col4_sum=('Col4', 'sum'))
         .reset_index())
print (df)
  Col1  Col2 Col3  Count  Col4_sum
0    A     1   AA      2        15
1    A     2   AB      1        30
2    B     4   FF      1        10
3    C     1   HH      1         4
4    C     3   GG      2        13
5    D     1   AA      1         4
6    D     3   FF      1         6

Answer 2

You can use a dict of column names and aggregation functions. See https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.aggregate.html

>>> df = pd.DataFrame([[1, 2, 3],
...                    [4, 5, 6],
...                    [7, 8, 9],
...                    [np.nan, np.nan, np.nan]],
...                   columns=['A', 'B', 'C'])
>>> df.agg({'A' : ['sum', 'min'], 'B' : ['min', 'max']})
#         A    B
# max   NaN  8.0
# min   1.0  2.0
# sum  12.0  NaN

Answer 3

Another solution that's a bit more verbose and hasn't been mentioned is to use the assign function as follows:

df = df1.assign(Count=df1.groupby(['Col1','Col2','Col3']).Col4.transform('size'))
        .assign(Col4_sum=df1.groupby(['Col1','Col2','Col3']).Col4.transform('sum'))
        .reset_index()

Answer 4

This should solve your problem.

df2 = df.groupby(['Col1','Col2','Col3'])['Col4'].agg('sum')

Answer 5

With the agg function and a dictionary, you can customise your output like so

df.groupby(['Col1','Col2','Col3']).agg({'Col3': ['count'], 'Col4': ['count','sum']})

This should return a group for Col1, Col2, and Col3, while aggregating the count for Col3, and then the count and sum for Col4

Answer 6

You can use the function pivot_table :

df = pd.pivot_table(df, index=['Col1', 'Col2', 'Col3'], values='Col4', aggfunc=['count', 'sum']).reset_index()
df.columns = ['Col1', 'Col2', 'Col3', 'Count', 'Col4_sum']

Output:

  Col1  Col2 Col3  Count  Col4_sum
0    A     1   AA      2        15
1    A     2   AB      1        30
2    B     4   FF      1        10
3    C     1   HH      1         4
4    C     3   GG      2        13
5    D     1   AA      1         4
6    D     3   FF      1         6