Categorical Encoding

One-Hot Encode Small

Create one binary column per category. For each row, the column for its category is 1 and all other columns are 0. By hand, build a per-category indicator list with a nested loop. With pandas, pd.get_dummies() does the same in one call — cast to int for 0/1 output since pandas returns bool by default.

By hand

With pandas

pd.get_dummies(df['c']) returns a DataFrame with one boolean column per category, in sorted alphabetical order. Chain .astype(int) to convert True/False to 1/0 — pandas 2.x returns bool by default. The snapshot shows dtype: int64 after the cast and the per-column indicator lists.

naive.py 
labels = ['cat', 'dog', 'bird', 'cat']
cats = sorted(set(labels))
result = {}
for c in cats:
    result[c] = []
for s in labels:
    for c in cats:
        result[c].append(1 if s == c else 0)
print('RESULT:', result)
library.py 
import pandas as pd
from dalib.display import set_display
set_display()

labels = ['cat', 'dog', 'bird', 'cat']
df = pd.DataFrame({'c': labels})
ohe = pd.get_dummies(df['c']).astype(int)
result = {c: ohe[c].tolist() for c in ohe.columns}
print('columns:', ohe.columns.tolist())
print('dtype:', ohe.dtypes[0])
print('RESULT:', result)

columns: ['bird', 'cat', 'dog']
dtype: int64
RESULT: {'bird': [0, 0, 1, 0], 'cat': [1, 0, 0, 1], 'dog': [0, 1, 0, 0]}

Implementation notes

  • pd.get_dummies() sorts columns alphabetically — the same order as sorted(set(labels)) in the naive half, so RESULT columns align.
  • The default dtype in pandas 2.x is bool. Add .astype(int) when downstream code expects 0/1 integers rather than True/False.
  • One-hot creates k columns for k categories; label-encode (this chapter) creates a single integer column — both represent the same information but one-hot avoids implying ordinal ranking.
  • Cross-reference: label-encode (this chapter) for the integer-code alternative.