Scikit-Learn GridSearchCV SelfTrainingClassifier

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 the SelfTrainingClassifier, a semi-supervised learning model.

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.

SelfTrainingClassifier is a semi-supervised learning model that extends a base classifier by iteratively adding high-confidence predictions to the training set. This allows the model to leverage both labeled and unlabeled data for training, improving its performance.

The key hyperparameters for SelfTrainingClassifier include the threshold, which is the confidence level for adding pseudo-labeled data; k_best, which specifies the number of high-confidence samples to add per iteration; and the criterion, which determines the stopping criteria for the self-training process.

from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.semi_supervised import SelfTrainingClassifier
from sklearn.tree import DecisionTreeClassifier

# Generate synthetic binary classification dataset
X, y = make_classification(n_samples=1000, n_features=10, n_informative=5, n_redundant=5, random_state=42)
y[::10] = -1  # Unlabel 10% of the data

# 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 parameter grid
param_grid = {
    'threshold': [0.7, 0.8, 0.9],
    'k_best': [10, 50, 100],
    'criterion': ['k_best', 'threshold']
}

# Perform grid search
base_classifier = DecisionTreeClassifier(random_state=42)
grid_search = GridSearchCV(estimator=SelfTrainingClassifier(base_classifier),
                           param_grid=param_grid,
                           cv=5,
                           scoring='accuracy')
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_
accuracy = best_model.score(X_test, y_test)
print(f"Test set accuracy: {accuracy:.3f}")

Running the example gives an output like:

Best score: 0.801
Best parameters: {'criterion': 'k_best', 'k_best': 100, 'threshold': 0.7}
Test set accuracy: 0.785

The steps are as follows:

1. Generate a synthetic binary classification dataset using scikit-learn’s make_classification function and unlabeled 10% of the data.
2. Split the dataset into train and test sets using train_test_split.
3. Define the parameter grid with different values for threshold, k_best, and criterion hyperparameters.
4. Perform grid search using GridSearchCV, specifying the SelfTrainingClassifier, parameter grid, 5-fold cross-validation, and accuracy scoring metric.
5. Report the best cross-validation score and best set of hyperparameters found by grid search.
6. Evaluate the best model on the hold-out test set and report the accuracy.

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 SelfTrainingClassifier model.