How to split a pandas dataframe based on column prefix

Question

I have a dataframe that I imported from a csv. It goes something like this:

df
A.1  B.1  A.2  B.2
1    1    1    1
2    2    2    2

My question is, what would be an efficient way to turn this into seperate data frames just comprised of A's and B's

df_a
A.1  A.2
1    1
2    2

df_b
B.1  B.2
1    1
2    2

I am not picky as far as the column names, would be fine with having them just be stripped to 1 and 2 etc but haven't been able to find a good way to do this. I am also open to other/better ways to accomplish what I am trying to do in case this doesn't make sense to someone more knowledgable. Thanks!

Answer 1

You could use df.filter with regex patterns:

df_a, df_B = df.filter(regex=r'^A'), df.filter(regex=r'^B')

or

df_a, df_B = df.filter(like='A'), df.filter(like='B')

Note that if you use like='A' then all columns whose name contains 'A' will be selected. If you use regex=r'^A' then only those columns whose name begins with an A will be selected.

---

In [7]: df
Out[7]: 
   A.1  B.1  A.2  B.2
0    1    1    1    1
1    2    2    2    2

In [8]: df_a, df_B = df.filter(regex=r'^A'), df.filter(regex=r'^B')

In [9]: df_a
Out[9]: 
   A.1  A.2
0    1    1
1    2    2

In [10]: df_B
Out[10]: 
   B.1  B.2
0    1    1
1    2    2

Answer 2

Ok, if I understand correctly you just need N new dataframes according to their column name.

dfa = df[[col for col in df.columns if col.startswith("A")]].copy()
# same for dfb, dfc...

Note that copy() is required if you wish to later apply changes to that new dataframe dfa . Otherwise, if I remember correctly, you would be applying changes by pointer, much like when using dictionaries.

Answer 3

To select the columns:

dfa = df[['A.1', 'A.2']]

To change the name of the columns:

dfa.reindex=["a1","a2"]