从数据框中随机选择列

Question

My question is quite simple: Is there any way to randomly choose columns from a dataframe in Pandas? To be clear, I want to randomly pick out n columns with the values attached. I know there is such a method for randomly picking rows:

import pandas as pd

df = pd.read_csv(filename, sep=',', nrows=None)
a = df.sample(n = 2)

So the question is, does it exist an equivalent method for finding random columns?

Answer 1

sample also accepts an axis parameter:

df = pd.DataFrame(np.random.randint(1, 10, (10, 5)), columns=list('abcde'))

df
Out: 
   a  b  c  d  e
0  4  5  9  8  3
1  7  2  2  8  7
2  1  5  7  9  2
3  3  3  5  2  4
4  8  4  9  8  6
5  6  5  7  3  4
6  6  3  6  4  4
7  9  4  7  7  3
8  4  4  8  7  6
9  5  6  7  6  9

df.sample(2, axis=1)
Out: 
   a  d
0  4  8
1  7  8
2  1  9
3  3  2
4  8  8
5  6  3
6  6  4
7  9  7
8  4  7
9  5  6

Answer 2

You can just do df.columns.to_series.sample(n=2)

to randomly sample the columns, first you need to convert to a Series by calling to_series then you can call sample as before

In[24]:
df.columns.to_series().sample(2)

Out[24]: 
C    C
A    A
dtype: object

Example:

In[30]:
df = pd.DataFrame(np.random.randn(5,3), columns=list('abc'))
df

Out[30]: 
          a         b         c
0 -0.691534  0.889799  1.137438
1 -0.949422  0.799294  1.360521
2  0.974746 -1.231078  0.812712
3  1.043434  0.982587  0.352927
4  0.462011 -0.591438 -0.214508

In[31]:
df[df.columns.to_series().sample(2)]

Out[31]: 
          b         a
0  0.889799 -0.691534
1  0.799294 -0.949422
2 -1.231078  0.974746
3  0.982587  1.043434
4 -0.591438  0.462011