I have a DataFrame column with 3 values - Bart, Peg, Human. I need to one-hot encode them such that Bart and Peg stay as columns and human is represented as 0 0.

Xi | Architecture
0  | Bart
1  | Bart
2  | Peg
3  | Human
4  | Human
5  | Peg
..
.

I want to one-hot encode them so that Human is represented as 0 0:

Xi |Bart| Peg
0  | 1  | 0
1  | 1  | 0
2  | 0  | 1
3  | 0  | 0
4  | 0  | 0
5  | 0  | 1

But when I do :

pd.get_dummies(df['Architecture'], drop_first = True)

it removes "Bart" and keeps the other 2. Is there a way to specify which column to remove?

CodePudding user response：

You could mask it:

df = df[['Xi']].join(pd.get_dummies(df['Architecture'].mask(df['Architecture']=='Human')))

Output:

   Xi  Bart  Peg
0   0     1    0
1   1     1    0
2   2     0    1
3   3     0    0
4   4     0    0
5   5     0    1

CodePudding user response：

IIUC, try use get_dummies then drop 'Human' column:

df['Architecture'].str.get_dummies().drop('Human', axis=1)

Output:

   Bart  Peg
0     1    0
1     1    0
2     0    1
3     0    0
4     0    0
5     0    1

CodePudding user response：

It's dropping "Bart" because that's the "first" label it sees. get_dummies doesn't have a built in way to say "drop this column after". It is annoying. So you can do a few things:

• sort the dataset before using get_dummies so "Human" shows up first when you use drop first
• subset the dataset to only one-hot-encode the columns where (architecture = "Bart" or "Peg")