CODEWITHSUNDEEP
pythonpython-3.xpandas
Krishh
Krishh

Reputation: 103

Pandas: a column with repeating groups to multiple columns

I have below pandas dataset:

set_id
A,B
A,C,E
A

Desired result:

set_id  set_id_1 set_id_2 set_id_3
A,B      A          B       null
A,C,E    A          C        E
A        A          null     null

set_id can have n number of values. Suppose the maximum number of values in set_id is 100, I should have 100 new columns

I tried using multilabel binarizer

df1 = pd.DataFrame()
df1['set_id'] = df['set_id'].str.split(',')
from sklearn.preprocessing import MultiLabelBinarizer  
mlb = MultiLabelBinarizer() 
df1=df.join(pd.DataFrame(mlb.fit_transform(df['set_id'])                                  ,columns=mlb.classes_,index=df.head(100).index))

It will create more than 100K columns as I have more than 100K unique records

Upvotes: 1

Views: 112

Answers (1)

jezrael
jezrael

Reputation: 862511

Use str.split with expand=True for DataFrame:

df1 = df['set_id'].str.split(',', expand=True)

Alternative faster solution with list comprehension:

df1 = pd.DataFrame([x.split(',') for x in df['set_id']])

df1.columns = [f'set_id_{x+1}' for x in df1.columns]
df1 = df.join(df1)

print (df1)
  set_id set_id_1 set_id_2 set_id_3
0    A,B        A        B     None
1  A,C,E        A        C        E
2      A        A     None     None

Upvotes: 1

Related Questions