Native Feature Names

Scikit-learn 1.0 introduces native feature name support via get_feature_names_out(). Transformers propagate column names through pipelines, eliminating the need to manually track mappings.

Pipeline Example

python 
import pandas as pd
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler, OneHotEncoder
from sklearn.compose import ColumnTransformer

df = pd.DataFrame({
    'age': [25, 30, 35],
    'salary': [30000, 50000, 70000],
    'city': ['Paris', 'Lyon', 'Paris'],
})

ct = ColumnTransformer([
    ('num', StandardScaler(), ['age', 'salary']),
    ('cat', OneHotEncoder(), ['city']),
])

ct.fit(df)
print(ct.get_feature_names_out())
# ['num__age', 'num__salary', 'cat__city_Lyon', 'cat__city_Paris']

Sources