CODEWITHSUNDEEP
pythonpandasdataframedata-manipulation
Tommy Yu
Tommy Yu

Reputation: 1104

Pandas: sum up every k columns

I have a dataframe ,how can I get column sum for every K cloumns for example, from

   Name    1       2         3       4
    A      2       3         5       8
    B      1       2         3       4
    C      1       6         8       9
    D      2       4         5       5
    E      3       3         3       3

summing for every 2 columns to

   Name   1       2   
    A     5       13 
    B     3       7  
    C     7       17   
    D     6       10  
    E     6       6

The window and number of columns can be very large. Is there any method to do that without the fot loop?

Upvotes: 1

Views: 121

Answers (2)

Ch3steR
Ch3steR

Reputation: 20669

You can use df.rolling over axis 1 with window size 2.

df.set_index('Name').rolling(2,axis=1).sum().iloc[:,1::2]

        2     4
Name
A     5.0  13.0
B     3.0   7.0
C     7.0  17.0
D     6.0  10.0
E     6.0   6.0

Upvotes: 2

jezrael
jezrael

Reputation: 862591

You can use integer division by // with helper array by length of columns and pass to groupby by axis=1 with aggregate sum:

#if Name is column
df = df.set_index('Name')

K = 2
df = df.groupby(np.arange(len(df.columns)) // K, axis=1).sum()
print (df)
      0   1
Name       
A     5  13
B     3   7
C     7  17
D     6  10
E     6   6

Details:

print (np.arange(len(df.columns)))
[0 1 2 3]

print (np.arange(len(df.columns)) // K)
[0 0 1 1]

Or if values are consecutive integers starting by 1 is possible subtract 1 and divide:

df = df.groupby((df.columns - 1) // K, axis=1).sum()
print (df)
      0   1
Name       
A     5  13
B     3   7
C     7  17
D     6  10
E     6   6

Upvotes: 1

Related Questions