Day 083 — Titanic攻略② — アンサンブルで上位25%を目指す

📚 背景知識（読んでから問題へ）

3つのモデルの役割分担

🗳️ Soft Voting の仕組み

ある乗客 X に対して3モデルが「生存確率」を予測した場合:

📋 アンサンブル手法比較

🎯 問題

💡 ヒント

from sklearn.ensemble import VotingClassifier, RandomForestClassifier, GradientBoostingClassifier
from sklearn.linear_model import LogisticRegression

lr  = LogisticRegression(max_iter=1000, C=1.0)
rf  = RandomForestClassifier(n_estimators=200, random_state=42)
gb  = GradientBoostingClassifier(n_estimators=200, learning_rate=0.05, random_state=42)

voting = VotingClassifier(
    estimators=[('lr', lr), ('rf', rf), ('gb', gb)],
    voting='soft'
)

from sklearn.model_selection import cross_val_score, StratifiedKFold
cv = StratifiedKFold(n_splits=5, shuffle=True, random_state=42)

for name, model in [('LR', lr), ('RF', rf), ('GB', gb), ('Voting', voting)]:
    scores = cross_val_score(model, X, y, cv=cv, scoring='accuracy')
    print(f"{name}: {scores.mean():.4f} ± {scores.std():.4f}")

✅ 模範解答

import pandas as pd
import numpy as np
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler, OneHotEncoder
from sklearn.compose import ColumnTransformer
from sklearn.impute import SimpleImputer
from sklearn.linear_model import LogisticRegression
from sklearn.ensemble import (RandomForestClassifier, GradientBoostingClassifier,
                               VotingClassifier)
from sklearn.model_selection import cross_val_score, StratifiedKFold

# ── 1. データ読み込み ──────────────────────────────
train = pd.read_csv('train.csv')
test  = pd.read_csv('test.csv')

# ── 2. 特徴量エンジニアリング（Day 082 の成果）────────
def engineer_features(df):
    df = df.copy()
    df['Title'] = df['Name'].str.extract(r',\s*([^\.]+)\.')
    rare = ['Jonkheer','Don','Sir','Capt','Col','Major','Rev',
            'Dr','Lady','Countess','Mme','Ms','Mlle','the Countess','Dona']
    df['Title'] = df['Title'].replace(rare, 'Rare')
    df['FamilySize'] = df['SibSp'] + df['Parch'] + 1
    df['IsAlone']    = (df['FamilySize'] == 1).astype(int)
    df['Deck']       = df['Cabin'].str[0].fillna('U')
    return df

train = engineer_features(train)
test  = engineer_features(test)

# ── 3. 特徴量選択 ──────────────────────────────────
features_num = ['Age', 'Fare', 'FamilySize']
features_cat = ['Pclass', 'Sex', 'Embarked', 'Title', 'Deck', 'IsAlone']

X      = train[features_num + features_cat]
y      = train['Survived']
X_test = test[features_num + features_cat]

# ── 4. 前処理パイプライン ─────────────────────────
num_pipe = Pipeline([
    ('impute', SimpleImputer(strategy='median')),
    ('scale',  StandardScaler()),
])
cat_pipe = Pipeline([
    ('impute', SimpleImputer(strategy='constant', fill_value='Missing')),
    ('ohe',    OneHotEncoder(handle_unknown='ignore', sparse_output=False)),
])
preprocessor = ColumnTransformer([
    ('num', num_pipe, features_num),
    ('cat', cat_pipe, features_cat),
])

# ── 5. モデル定義 ────────────────────────────────
lr  = LogisticRegression(max_iter=1000, C=1.0, random_state=42)
rf  = RandomForestClassifier(n_estimators=200, max_depth=6,
                              min_samples_leaf=4, random_state=42)
gb  = GradientBoostingClassifier(n_estimators=200, learning_rate=0.05,
                                  max_depth=3, random_state=42)
voting = VotingClassifier(
    estimators=[('lr', lr), ('rf', rf), ('gb', gb)],
    voting='soft',
    weights=[1, 2, 2],
)

# ── 6. CV評価 ──────────────────────────────────────
cv = StratifiedKFold(n_splits=5, shuffle=True, random_state=42)

results = {}
for name, base_model in [('LR', lr), ('RF', rf), ('GB', gb), ('Voting', voting)]:
    pipe   = Pipeline([('prep', preprocessor), ('model', base_model)])
    scores = cross_val_score(pipe, X, y, cv=cv, scoring='accuracy')
    results[name] = scores
    print(f"{name:8s}: {scores.mean():.4f} ± {scores.std():.4f}")

# ── 7. 最終モデルで全訓練データ学習 → 提出ファイル ─────
final_pipe = Pipeline([('prep', preprocessor), ('model', voting)])
final_pipe.fit(X, y)

preds      = final_pipe.predict(X_test)
submission = pd.DataFrame({'PassengerId': test['PassengerId'], 'Survived': preds})
submission.to_csv('submission.csv', index=False)
print("\nsubmission.csv を生成しました！")
print(submission['Survived'].value_counts())

🔢 Step-by-Step 解説

VotingClassifier(
    estimators=[('lr', lr), ('rf', rf), ('gb', gb)],
    voting='soft',
    weights=[1, 2, 2],
)

📈 精度比較チャート（期待値）

📐 数学補足（文系向け）

Soft Voting の計算式

なぜアンサンブルが効くのか（分散の削減）

🏅 Kaggleでの実践的な使い方

⚠️ よくある誤解・ミス

🚀 次のステップ

📝 自己評価（解いた後に記入）