🔒 05 — Pipelines and Leakage¶

Pipelines keep preprocessing and modeling together.

ColumnTransformer¶

from sklearn.compose import ColumnTransformer
from sklearn.preprocessing import OneHotEncoder, StandardScaler
from sklearn.pipeline import Pipeline
from sklearn.linear_model import LogisticRegression

numeric_features = ["age", "income"]
categorical_features = ["city", "segment"]

preprocess = ColumnTransformer([
    ("num", StandardScaler(), numeric_features),
    ("cat", OneHotEncoder(handle_unknown="ignore"), categorical_features)
])

model = Pipeline([
    ("preprocess", preprocess),
    ("model", LogisticRegression(max_iter=1000))
])

model.fit(X_train, y_train)

Why This Prevents Leakage¶

When used correctly:

scalers fit only on training data
encoders learn categories from training data
test data is transformed using training rules

Common Leakage Examples¶

filling missing values before split using full data
scaling before split
creating target-based features on full data
using future information

Next¶

➡️ 06-exercises