python - is it possible to concat column after using pandas get_dummies?

Question

here is my example df

         doc_num
doc1 doc2 
 A    B    U123
 A    C    U123
 A    D    U124
 B    C    U126
 B    D    U126

and i have use

pd.get_dummies(df.doc_num).sort_index(level=0)

to make a vector matrix like this

           U123 U124 U126
doc1 doc2  
 A    B     1    0    0
 A    C     1    0    0
 A    D     0    1    0
 B    C     0    0    1
 B    D     0    0    1

but i would like to concat the doc1 and doc2 then create a new column to see the expected result like this

       U123 U124 U126
doc_3  
 A,B     1    0    0
 A,C     1    0    0
 A,D     0    1    0
 B,C     0    0    1
 B,D     0    0    1

is it possible? thank you in advance

Answer 1

In addition to @jezrael's answer, you want a vector matrix, so do:

df1=pd.get_dummies(df.doc_num)
df1.insert(0, 'doc_3',  df['doc1'] + ',' + df['doc2'])
print(df1.set_index('doc_3'))

Or:

df1=pd.get_dummies(df.doc_num)
df1['doc_3']=df.pop('doc1') + ',' + df.pop('doc2')
print(df1.set_index('doc_3'))

All Output:

       U123  U124  U126
doc_3                  
A,B       1     0     0
A,C       1     0     0
A,D       0     1     0
B,C       0     0     1
B,D       0     0     1

Now you really get your desired output.

Answer 2

I believe you need join both levels of MultiIndex , set index name by rename_axis :

df1 = pd.get_dummies(df.doc_num).sort_index(level=0)
df1.index = df1.index.map(','.join)
df1 = df1.rename_axis('doc_3')
print (df1)
       U123  U124  U126
doc_3                  
A,B       1     0     0
A,C       1     0     0
A,D       0     1     0
B,C       0     0     1
B,D       0     0     1

And add reset_index for column if necessary:

df1 = df1.reset_index()
print (df1)
  doc_3  U123  U124  U126
0   A,B     1     0     0
1   A,C     1     0     0
2   A,D     0     1     0
3   B,C     0     0     1
4   B,D     0     0     1

---

Or first reset_index to columns from MultiIndex with pop for extract columns if want index:

df1 = pd.get_dummies(df.doc_num).sort_index(level=0).reset_index()
df1.index =  df1.pop('doc1') + ',' + df1.pop('doc2')
df1 = df1.rename_axis('doc_3')
print (df1)
       U123  U124  U126
doc_3                  
A,B       1     0     0
A,C       1     0     0
A,D       0     1     0
B,C       0     0     1
B,D       0     0     1

---

Or use insert for new column:

df1 = pd.get_dummies(df.doc_num).sort_index(level=0).reset_index()
df1.insert(0, 'doc_3',  df1.pop('doc1') + ',' + df1.pop('doc2'))

print (df1)
  doc_3  U123  U124  U126
0   A,B     1     0     0
1   A,C     1     0     0
2   A,D     0     1     0
3   B,C     0     0     1
4   B,D     0     0     1

Answer 3

You can try below code. It will combine two columns into one . Also, add "," in between them.

df['doc_3'] = df['doc1'] + "," + df['doc2']

Then you can drop first two columns