用于嵌套表的 Pandas Get_dummies

Question

I am looking to utilize pandas get_dummy() functionality to encode a (quite extensive) set of categorical variables. However the data is currently in nested table format. Meaning that each row represents another variable instance for example

Instance, Cat_Col
1, John
1, Smith
2, Jane
3, Joe

Now I can generate the full list of unique variables which I can use to get_dummies which represent all possible values. However transforming the nested table into a single instance row in this new format is giving me some trouble.

Any help is much appreciated Thanks

Edit: each instance should have a dummy coding result for all values of Cat_col

The idea would be the result be a single feature vector like so

Instance,Col_John,Col_Smith,Col_Jane,Col_Joe
1,1,1,0,0
2,0,0,1,0
3,0,0,0,1

I believe that is the correct coding, assuming we are doing 1-hot encoding

Answer 1

You may want to consider using pivot_table to achieve your goal here.

import pandas as pd

df

Out[10]: 
   Instance Cat_Col
0         1    John
1         1   Smith
2         2    Jane
3         3     Joe

df['count'] = 1
df.pivot('Instance', 'Cat_Col', 'count').fillna(0)

Out[11]: 
Cat_Col    Jane   Joe   John   Smith
Instance                            
1             0     0      1       1
2             1     0      0       0
3             0     1      0       0

If you prefer to use get_dummies ,

result = pd.get_dummies(df.Cat_Col)
result['Instance'] = df.Instance
result = result.set_index('Instance')
result.groupby(level=0).apply(max)

Out[26]: 
           Jane   Joe   John   Smith
Instance                            
1             0     0      1       1
2             1     0      0       0
3             0     1      0       0