正解率（Accuracy）| Python 3.13 での基礎と落とし穴

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
import numpy as np
from sklearn.datasets import load_breast_cancer
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score, balanced_accuracy_score
from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline, make_pipeline
from sklearn.preprocessing import StandardScaler


def build_model(random_state: int = 42) -> Pipeline:
    """乳がん診断データを分類するランダムフォレストのパイプラインを構築する。"""
    return make_pipeline(
        StandardScaler(),
        RandomForestClassifier(random_state=random_state, n_estimators=300),
    )


def plot_accuracy_comparison() -> None:
    """Accuracy と Balanced Accuracy を比較する棒グラフを生成・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=0.25,
        stratify=labels,
        random_state=42,
    )

    pipeline = build_model()
    pipeline.fit(x_train, y_train)
    predictions = pipeline.predict(x_test)

    acc = accuracy_score(y_test, predictions)
    bal_acc = balanced_accuracy_score(y_test, predictions)
    print(f"Accuracy: {acc:.3f}, Balanced Accuracy: {bal_acc:.3f}")

    figure, axis = plt.subplots(figsize=(5, 4))
    scores = np.array([acc, bal_acc])
    labels = ["Accuracy", "Balanced Accuracy"]
    colors = ["#2563eb", "#f97316"]

    axis.bar(labels, scores, color=colors)
    axis.set_ylim(0, 1.05)
    axis.set_ylabel("Score")
    axis.set_title("Accuracy vs. Balanced Accuracy (Breast Cancer Dataset)")
    axis.grid(axis="y", linestyle="--", alpha=0.4)

    for label, score in zip(labels, scores, strict=True):
        axis.text(label, score + 0.02, f"{score:.3f}", ha="center", va="bottom")

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


if __name__ == "__main__":
    plot_accuracy_comparison()