Group by one columns and find sum and max value for another in pandas

问题

I have a dataframe like this:

Name  id  col1  col2  col3  cl4 
PL    252  0     747   3     53  
PL2   252  1     24    2     35 
PL3   252  4     75    24    13 
AD    889  53    24    0     95 
AD2   889  23    2     0     13  
AD3   889  0     24    3     6  
BG    024  12    89    53    66 
BG1   024  43    16    13    0   
BG2   024  5     32    101   4

And now I need to group by ID, and for columns col1 and col4 find the sum for each id and put that into a new column near to parent column (example: col3(sum)) But for col2 and col3 find max value. Desired output:

Name  id  col1 col1(sum) col2 col2(max) col3 col(max) col4 col4(sum)
PL    252  0       5      747    747     3     24    6    18
PL2   252  1       5      24     747     2     24    12   18
PL3   252  4       5      75     747     24    24    0    18
AD    889  53      76     24     24      95    95    23   33
AD2   889  23      76     2      24      13    95    5    33
AD3   889  0       76     24     24      6     95    5    33
BG    024  12      60     89     89      66    66    0    67   
BG1   024  43      60     16     89      0     66    63   67    
BG2   024  5       60     32     89      4     66    4    67

What is the easiest and fastest way to calculate this?

回答1:

You can use groupby/transform to creat the required columns

df[['col1_sum', 'col4_sum']]=df.groupby('id')['col1', 'cl4'].transform('sum')
df[['col2_max', 'col3_max']]=df.groupby('id')['col1', 'cl4'].transform('max')

    Name    id  col1    col2    col3    cl4 col1_sum    col4_sum    col2_max    col3_max
0   PL      252 0       747     3       53  5           101         4   53
1   PL2     252 1       24      2       35  5           101         4   53
2   PL3     252 4       75      24      13  5           101         4   53
3   AD      889 53      24      0       95  76          114         53  95
4   AD2     889 23      2       0       13  76          114         53  95
5   AD3     889 0       24      3       6   76          114         53  95
6   BG      24  12      89      53      66  60          70          43  66
7   BG1     24  43      16      13      0   60          70          43  66
8   BG2     24  5       32      101     4   60          70          43  66

回答2:

The most (pandas) native way to do this, is to use the .agg() method that allows you to specify the aggregation function you want to apply per column (just like you would do in SQL).

Sample from the documentation:

df.groupby('A').agg({'B': ['min', 'max'], 'C': 'sum'})

回答3:

You can use merge when you have groupby and sum on id :

pd.merge(df,df.groupby("id").sum().reset_index(), on='id',how='outer')

output

回答4:

I know this is messy but I like chaining so you can do something like this:

df = df.groupby('id').
    apply(lambda g: g.assign(
        col1_sum=g.col1.sum(),
        col2_max=g.col2.max()))

Basically, this is applying a group based assign command to each group and then combining into a single DataFrame.

See https://pandas.pydata.org/pandas-docs/stable/api.html for details on each method.

来源：https://stackoverflow.com/questions/44724480/group-by-one-columns-and-find-sum-and-max-value-for-another-in-pandas

标签

python

pandas

dataframe

group-by