Scikit-Learn GridSearchCV StackingRegressor

Hyperparameter tuning is a crucial step in optimizing machine learning models for best performance. In this example, we’ll demonstrate how to use scikit-learn’s GridSearchCV to perform hyperparameter tuning for StackingRegressor, an ensemble model that combines multiple regression models to improve predictive performance.

Grid search is a method for evaluating different combinations of model hyperparameters to find the best performing configuration. It exhaustively searches through a specified parameter grid, trains and evaluates the model for each combination using cross-validation, and selects the hyperparameters that yield the best performance metric.

The stacking regressor combines several base estimators to form a robust ensemble model. It uses a meta-regressor to combine the predictions of the base estimators, which can lead to better performance compared to individual models.

Key hyperparameters for the stacking regressor include the choice of base estimators, the final estimator, and parameters specific to each base estimator and the final estimator. Tuning these hyperparameters helps in achieving optimal performance.

from sklearn.datasets import make_regression
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.linear_model import Ridge, Lasso
from sklearn.ensemble import RandomForestRegressor, StackingRegressor
from sklearn.svm import SVR

# Generate synthetic regression dataset
X, y = make_regression(n_samples=1000, n_features=20, noise=0.1, random_state=42)

# Split into train and test sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Define base estimators
base_estimators = [
    ('ridge', Ridge()),
    ('lasso', Lasso()),
    ('svr', SVR())
]

# Define parameter grid
param_grid = {
    'ridge__alpha': [0.1, 1, 10],
    'lasso__alpha': [0.1, 1, 10],
    'svr__C': [0.1, 1, 10],
    'final_estimator': [RandomForestRegressor(n_estimators=10, random_state=42)]
}

# Define Stacking Regressor
stacking_regressor = StackingRegressor(estimators=base_estimators, final_estimator=RandomForestRegressor())

# Perform grid search
grid_search = GridSearchCV(estimator=stacking_regressor,
                           param_grid=param_grid,
                           cv=5,
                           scoring='r2')
grid_search.fit(X_train, y_train)

# Report best score and parameters
print(f"Best score: {grid_search.best_score_:.3f}")
print(f"Best parameters: {grid_search.best_params_}")

# Evaluate on test set
best_model = grid_search.best_estimator_
r2_score = best_model.score(X_test, y_test)
print(f"Test set R-squared: {r2_score:.3f}")

Running the example gives an output like:

Best score: 1.000
Best parameters: {'final_estimator': RandomForestRegressor(n_estimators=10, random_state=42), 'lasso__alpha': 0.1, 'ridge__alpha': 10, 'svr__C': 1}
Test set R-squared: 1.000

The steps are as follows:

1. Generate a synthetic regression dataset using scikit-learn’s make_regression function.
2. Split the dataset into train and test sets using train_test_split.
3. Define the parameter grid with different values for ridge__alpha, lasso__alpha, svr__C, and the final estimator.
4. Perform grid search using GridSearchCV, specifying the StackingRegressor, parameter grid, 5-fold cross-validation, and R-squared scoring metric.
5. Report the best cross-validation score and the optimal set of hyperparameters found by grid search.
6. Evaluate the best model on the hold-out test set and report the R-squared score.

By using GridSearchCV, we can easily explore different hyperparameter settings and find the combination that maximizes the model’s performance. This automated approach saves time and effort compared to manual hyperparameter tuning and helps ensure we select the best configuration for our stacking regressor model.