plot_roc_curves.py

from sklearn import datasets
from sklearn.tree import DecisionTreeClassifier
from sklearn import model_selection
from sklearn.metrics import roc_curve, roc_auc_score
import matplotlib.pyplot as plt
from sklearn.model_selection import GridSearchCV

def plot_roc_curve_and_auc(dataset_id, criterion):
    # Load dataset
    dataset = datasets.fetch_openml(data_id=dataset_id)
    
    # Decision tree with specified criterion
    decision_tree = DecisionTreeClassifier(criterion=criterion)
    decision_tree.fit(dataset.data, dataset.target)

    # Plot ROC curve and compute AUC value
    plt.figure(figsize=(8, 6))

    # Decision tree with specified criterion
    y_scores = model_selection.cross_val_predict(decision_tree, dataset.data, dataset.target, method="predict_proba", cv=10)
    fpr, tpr, _ = roc_curve(dataset.target, y_scores[:, 1], pos_label="1")
    auc = roc_auc_score(dataset.target, y_scores[:, 1])
    plt.plot(fpr, tpr, label=f"{criterion.capitalize()} (AUC = {auc:.2f})")

    # Plot settings
    plt.plot([0, 1], [0, 1], color='gray', linestyle='--')
    plt.xlabel('1 - Specificity')
    plt.ylabel('Sensitivity')
    plt.title(f'ROC Curve - {criterion.capitalize()} - Dataset ID {dataset_id}')
    plt.legend()
    plt.grid(True)
    plt.savefig(f"roc_curve_{dataset_id}_{criterion}.png", dpi=150, bbox_inches="tight")
    plt.show()

    # Print AUC value
    print(f"AUC ({criterion.capitalize()}): {auc:.2f}")

    # Parameter search on min_samples_leaf
    parameters = [{"min_samples_leaf": [1, 10, 20, 50, 100]}]
    tuned_dtc = GridSearchCV(decision_tree, parameters, scoring="roc_auc", cv=10)
    tuned_dtc.fit(dataset.data, dataset.target)
    print(f"Best parameters for {criterion} criterion:", tuned_dtc.best_params_)

if __name__ == "__main__":
    plot_roc_curve_and_auc(4534, "entropy")
    plot_roc_curve_and_auc(4534, "gini")
    plot_roc_curve_and_auc(44, "entropy")
    plot_roc_curve_and_auc(44, "gini")