Tutorial 08: scikit-learn Pipelines

Level: Intermediate · Part of: ML Learning Roadmap

← Back to roadmap

Case Study: Housing Price Prediction Workflow

Scenario

A real estate analytics team needs a reproducible workflow that chains scaling, imputation, and regression — preventing data leakage and simplifying deployment.

Learning Objectives

• Build a Pipeline with multiple steps
• Access intermediate steps with named_steps
• Use make_pipeline shorthand
• Serialize with joblib

Prerequisites

• Tutorials 05–07
• Python 3.9+, scikit-learn, NumPy, pandas, matplotlib

Dataset

fetch_california_housing() subset

Hands-On Solution

Copy and run the complete script below:

from sklearn.datasets import fetch_california_housing
from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.impute import SimpleImputer
from sklearn.linear_model import Ridge
from sklearn.metrics import mean_squared_error
import joblib

X, y = fetch_california_housing(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

pipe = Pipeline([
    ("imputer", SimpleImputer(strategy="median")),
    ("scaler", StandardScaler()),
    ("model", Ridge(alpha=1.0)),
])

pipe.fit(X_train, y_train)
rmse = mean_squared_error(y_test, pipe.predict(X_test), squared=False)
print(f"RMSE: ${rmse * 100_000:,.0f}")  # prices in $100k units

# Save entire pipeline
joblib.dump(pipe, "housing_pipeline.joblib")
loaded = joblib.load("housing_pipeline.joblib")
print(f"Loaded pipeline score: {loaded.score(X_test, y_test):.4f}")

Expected Output

When you run the script, you should see evaluation metrics printed to the console. Some tutorials also save .png plot files in the current directory.

Exercises

1. Add a second model step using GridSearchCV on alpha (preview for tutorial 23).
2. Inspect pipe.named_steps['model'].coef_ after fitting.
3. Replace Ridge with ElasticNet. How does RMSE change?

Key Takeaways

• Pipelines bundle preprocessing + model into one estimators
• One fit() call handles everything in order
• Serialize the whole pipeline for production deployment

Navigation

Part of the ML Learning Roadmap — Hands-On with scikit-learn