根据列名拆分pandas数据框

Question

Is there a way to split a pandas data frame based on the column name? As an example consider the data frame has the following columns df = ['A_x', 'B_x', 'C_x', 'A_y', 'B_y', 'C_y'] and I want to create two data frames X = ['A_x', 'B_x', 'C_x'] and Y = ['A_y', 'B_y', 'C_y'] .

I know there is a possibility to do this:

d = {'A': df.A_x, 'B': df.B_x, 'C': df.B_x}
X = pd.DataFrame (data=d)

but this would not be ideal as in my case I have 2200 columns in df . Is there a more elegant solution?

Answer 1

You could use df.filter(regex=...) :

import numpy as np
import pandas as pd
df = pd.DataFrame(np.random.randn(2, 10),
                  columns='Time A_x A_y A_z B_x B_y B_z C_x C_y C-Z'.split())
X = df.filter(regex='_x')
Y = df.filter(regex='_y')

yields

In [15]: X
Out[15]: 
        A_x       B_x       C_x
0 -0.706589  1.031368 -0.950931
1  0.727826  0.879408 -0.049865

In [16]: Y
Out[16]: 
        A_y       B_y       C_y
0 -0.663647  0.635540 -0.532605
1  0.326718  0.189333 -0.803648