予測モデルの選定！ホールドアウトとクロスバリデーションの実践手順

import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets import fetch_california_housing
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LinearRegression
from sklearn.svm import SVR
from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import RandomForestRegressor, GradientBoostingRegressor
from sklearn.metrics import r2_score
from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import Pipeline


plt.rcParams['font.family'] = 'Hiragino sans'
# データ読み込み
data = fetch_california_housing()
X, y = data.data, data.target

# 訓練データ・テストデータに分割
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# モデル一覧（必要なものはStandardScalerで前処理）
models = {
    '線形回帰モデル': Pipeline([
        ('scaler', StandardScaler()),
        ('model', LinearRegression())
    ]),
    'SVM (RBF)': Pipeline([
        ('scaler', StandardScaler()),
        ('model', SVR(kernel='rbf'))
    ]),
    '決定木': DecisionTreeRegressor(random_state=42),
    'ランダムフォレスト': RandomForestRegressor(random_state=42),
    'Gradient Boosting': GradientBoostingRegressor(random_state=42)
}

# プロット準備
plt.figure(figsize=(18, 10))

# 各モデルで処理
for i, (name, model) in enumerate(models.items(), 1):
    model.fit(X_train, y_train)

    # 予測
    y_train_pred = model.predict(X_train)
    y_test_pred = model.predict(X_test)

    # R²スコア
    r2_train = r2_score(y_train, y_train_pred)
    r2_test = r2_score(y_test, y_test_pred)
    print(f"🔍 {name}")
    print(f"  訓練データ R²: {r2_train:.3f}")
    print(f"  テストデータ R²: {r2_test:.3f}\n")

    # 散布図
    plt.subplot(2, 3, i)
    plt.scatter(y_train, y_train_pred, color='blue', alpha=0.4, label='Train')
    plt.scatter(y_test, y_test_pred, color='red', alpha=0.4, label='Test')
    plt.plot([y.min(), y.max()], [y.min(), y.max()], 'k--', lw=2)
    plt.title(name)
    plt.xlabel("実測値")
    plt.ylabel("予測値")
    plt.legend()

plt.tight_layout()
plt.savefig("california_housing_regression.webp",format="webp",dpi=300)
plt.show()

import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets import fetch_california_housing
from sklearn.model_selection import cross_val_score, cross_val_predict, KFold
from sklearn.linear_model import LinearRegression
from sklearn.svm import SVR
from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import RandomForestRegressor, GradientBoostingRegressor
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.metrics import r2_score

# データ取得
data = fetch_california_housing()
X, y = data.data, data.target

# モデルの定義（スケーリングが必要なものはPipelineに）
models = {
    '線形回帰モデル': Pipeline([
        ('scaler', StandardScaler()),
        ('model', LinearRegression())
    ]),
    'SVM (RBF)': Pipeline([
        ('scaler', StandardScaler()),
        ('model', SVR(kernel='rbf'))
    ]),
    '決定木': DecisionTreeRegressor(random_state=42),
    'ランダムフォレスト': RandomForestRegressor(random_state=42),
    'Gradient Boosting': GradientBoostingRegressor(random_state=42)
}

# クロスバリデーション設定
cv = KFold(n_splits=5, shuffle=True, random_state=42)

# プロット準備
plt.figure(figsize=(18, 10))

# モデルごとに評価＆可視化
for i, (name, model) in enumerate(models.items(), 1):
    # CVスコア（R^2）
    scores = cross_val_score(model, X, y, cv=cv, scoring='r2')
    print(f"🔍 {name}")
    print(f"  R^2 平均: {scores.mean():.3f}, 標準偏差: {scores.std():.3f}")

    # CV予測
    y_pred = cross_val_predict(model, X, y, cv=cv)

    # 全体R^2
    overall_r2 = r2_score(y, y_pred)
    print(f"  cross_val_predict による全体R^2: {overall_r2:.3f}\n")

    # 散布図プロット
    plt.subplot(2, 3, i)
    plt.scatter(y, y_pred, alpha=0.4, label='CV Predictions')
    plt.plot([y.min(), y.max()], [y.min(), y.max()], 'k--', lw=2)
    plt.title(f"{name}\nR^2: {overall_r2:.3f}")
    plt.xlabel("実測値")
    plt.ylabel("予測値")
    plt.tight_layout()

plt.suptitle("クロスバリデーションによるモデル選定", fontsize=16, y=1.02)
plt.savefig("california_housing_regression_cv.webp",format="webp",dpi=300)
plt.show()