How to create multiple diff columns in pandas? At once

user_12845199 0
Viewed 54

Check the following dataset:

d = {'col1':[2,4,5,6],'col2':[8,8,6,1],'col3':[1,2,3,4],'col4':[4,4,4,4]}

df = pd.DataFrame(d,index=['Item1','Item2','Item1','Item2'])

enter image description here

Lets say i want to interact through the columns, and create a diff column. That gives me difference beetween col1, col2 and them another diff column that gives me col3,col4. And so on like in a window by window style. Difference columns.

d = {'col1':[2,4,5,6],'col2':[8,8,6,1],'diff_1':[-6,-4,1,5],'col3':[1,2,3,4],'col4':[4,4,4,4],'diff_2':[-3,-2,-1,0]}
    
df = pd.DataFrame(d,index=['Item1','Item2','Item1','Item2'])

enter image description here

pandas numpy python
3 Answers

You can check groupby then concat back

out = pd.concat([y.assign(**{'diff_{0}'.format(x+1): y.iloc[:,0] - y.iloc[:,1]})for x , y in df.groupby(np.arange(df.shape[1])//2,axis=1)],axis=1)
Out[166]: 
       col1  col2  diff_1  col3  col4  diff_2
Item1     2     8      -6     1     4      -3
Item2     4     8      -4     2     4      -2
Item1     5     6      -1     3     4      -1
Item2     6     1       5     4     4       0
user_79645270

define a chunk function which iterate through a list

def chunk_list(lst, chunk_size):
    for j in range(0, len(lst), chunk_size):
        yield lst[j:j + chunk_size]

then, use it as follows:

import pandas as pd
d = {'col1':[2,4,5,6],'col2':[8,8,6,1],'col3':[1,2,3,4],'col4':[4,4,4,4]}

df = pd.DataFrame(d,index=['Item1','Item2','Item1','Item2'])

for chunk in chunk_list(df.columns, 2):
    diff = "diff" + chunk[0]+ chunk[1]
    df[diff] = df[chunk[0]] - df[chunk[1]]
user_141313150

The idea below is, that you subtract all consecutive columns and then just pick the 2nd and the 4th diff columns:

# Use df.diff(axis=1) to subtract consecutive columns with each other. The multiply the result with -1
In [1575]: df_diff = df.diff(axis=1).mul(-1)[['col2', 'col4']].rename(columns={'col2': 'diff_1', 'col4': 'diff_2'})

# Concat the above result with the original dataframe
In [1577]: res = pd.concat([df, df_diff], axis=1)

In [1578]: res
Out[1578]: 
       col1  col2  col3  col4  diff_1  diff_2
Item1     2     8     1     4      -6      -3
Item2     4     8     2     4      -4      -2
Item1     5     6     3     4      -1      -1
Item2     6     1     4     4       5       0

Explanation:

In [1579]: df.diff(axis=1)
Out[1579]: 
       col1  col2  col3  col4
Item1   NaN     6    -7     3
Item2   NaN     4    -6     2
Item1   NaN     1    -3     1
Item2   NaN    -5     3     0

The above result is basically like:

col1 = col1 - Nan # You don't need this
col2 = col2 - col1 # You need col1 - col2. So mul this by -1
col3 = col3 - col2 # You don't need this
col4 = col4 - col3 # You need col3 - col4. So mul this by -1

So below command produces the required diff columns.

df_diff = df.diff(axis=1).mul(-1)[['col2', 'col4']].rename(columns={'col2': 'diff_1', 'col4': 'diff_2'})
user_58208140
Related