ROC-AUC | 閾値設計とモデル比較を支える指標

python --version        # 例: Python 3.13.0
pip install scikit-learn matplotlib

from __future__ import annotations

from pathlib import Path

import matplotlib.pyplot as plt
from sklearn.datasets import load_breast_cancer
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import RocCurveDisplay, roc_auc_score
from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline, make_pipeline
from sklearn.preprocessing import StandardScaler


def build_pipeline(random_state: int = 42) -> Pipeline:
    """ロジスティック回帰で ROC 曲線を計算するためのパイプラインを構築する。"""
    return make_pipeline(
        StandardScaler(),
        LogisticRegression(max_iter=2000, solver="lbfgs", random_state=random_state),
    )


def plot_roc_curve() -> None:
    """ROC 曲線と AUC を計算し、可視化してファイルに保存する。"""
    features, labels = load_breast_cancer(return_X_y=True)
    x_train, x_test, y_train, y_test = train_test_split(
        features,
        labels,
        test_size=0.3,
        stratify=labels,
        random_state=42,
    )

    pipeline = build_pipeline()
    pipeline.fit(x_train, y_train)
    probabilities = pipeline.predict_proba(x_test)[:, 1]

    auc = roc_auc_score(y_test, probabilities)
    print(f"ROC-AUC: {auc:.3f}")

    figure, axis = plt.subplots(figsize=(5, 5))
    RocCurveDisplay.from_predictions(
        y_test,
        probabilities,
        name="Logistic Regression",
        ax=axis,
    )
    axis.plot([0, 1], [0, 1], "--", color="grey", alpha=0.5, label="Random")
    axis.set_xlabel("False Positive Rate")
    axis.set_ylabel("True Positive Rate")
    axis.set_title("ROC Curve (Breast Cancer Dataset)")
    axis.legend(loc="lower right")

    figure.tight_layout()
    output_dir = Path("static/images/eval/classification/rocauc")
    output_dir.mkdir(parents=True, exist_ok=True)
    figure.savefig(output_dir / "roc_curve.png", dpi=150)
    plt.close(figure)


if __name__ == "__main__":
    plot_roc_curve()