演習：Telco Churnデータを徹底分析しよう

import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

df = pd.read_csv('WA_Fn-UseC_-Telco-Customer-Churn.csv')

# 1. データ概要
print(f"データサイズ: {df.shape}")
print(f"\nデータ型:\n{df.dtypes}")

# 2. 基本統計量
print(f"\n数値変数の統計量:\n{df.describe()}")
for col in df.select_dtypes(include='object').columns:
    print(f"\n{col}: {df[col].nunique()}種類")
    print(df[col].value_counts())

# 3. 欠損値
df['TotalCharges'] = pd.to_numeric(df['TotalCharges'], errors='coerce')
print(f"\n欠損値:\n{df.isnull().sum()[df.isnull().sum() > 0]}")
# 対処: tenure=0の新規顧客 → TotalCharges=0で補完

# 4. 目的変数の分布
print(f"\n離反分布:\n{df['Churn'].value_counts()}")
print(f"離反率: {(df['Churn'] == 'Yes').mean():.3f}")

# 5. カテゴリカル変数の可視化
cat_cols = df.select_dtypes(include='object').columns.drop(['customerID', 'Churn'])
fig, axes = plt.subplots(3, 4, figsize=(20, 12))
axes = axes.flatten()
for i, col in enumerate(cat_cols):
    df[col].value_counts().plot(kind='bar', ax=axes[i])
    axes[i].set_title(col)
    axes[i].tick_params(axis='x', rotation=45)
for j in range(i+1, len(axes)):
    axes[j].set_visible(False)
plt.tight_layout()
plt.savefig('categorical_distributions.png', dpi=150)

import pandas as pd
import matplotlib.pyplot as plt
from scipy import stats

df = pd.read_csv('WA_Fn-UseC_-Telco-Customer-Churn.csv')
df['TotalCharges'] = pd.to_numeric(df['TotalCharges'], errors='coerce').fillna(0)

results = []

# H1: Month-to-month契約は離反率が高い
h1_rates = df.groupby('Contract')['Churn'].apply(lambda x: (x=='Yes').mean())
results.append({
    '仮説': 'H1: Month-to-month契約は離反率が高い',
    '結果': f"M-t-M: {h1_rates['Month-to-month']:.1%}, 1yr: {h1_rates['One year']:.1%}, 2yr: {h1_rates['Two year']:.1%}",
    '判定': '支持',
    '示唆': '長期契約への移行インセンティブが有効'
})

# H2: 月額料金が高い顧客は離反しやすい
churned = df[df['Churn']=='Yes']['MonthlyCharges']
retained = df[df['Churn']=='No']['MonthlyCharges']
t_stat, p_val = stats.ttest_ind(churned, retained)
results.append({
    '仮説': 'H2: 月額料金が高い顧客は離反しやすい',
    '結果': f"離反者中央値: ¥{churned.median():.0f}, 非離反: ¥{retained.median():.0f}, p={p_val:.2e}",
    '判定': '支持',
    '示唆': '高額プラン顧客に割引や付加価値の提供を検討'
})

# H4: Fiber optic顧客は離反率が高い
h4_rates = df.groupby('InternetService')['Churn'].apply(lambda x: (x=='Yes').mean())
results.append({
    '仮説': 'H4: Fiber optic顧客は離反率が高い',
    '結果': f"Fiber: {h4_rates['Fiber optic']:.1%}, DSL: {h4_rates['DSL']:.1%}",
    '判定': '支持',
    '示唆': 'Fiber opticの品質/価格に対する不満を調査'
})

# H7: 利用期間が短い顧客は離反しやすい
df['tenure_group'] = pd.cut(df['tenure'], bins=[0,12,24,48,72], labels=['0-12','13-24','25-48','49-72'])
h7_rates = df.groupby('tenure_group')['Churn'].apply(lambda x: (x=='Yes').mean())
results.append({
    '仮説': 'H7: 利用期間が短い顧客は離反しやすい',
    '結果': f"0-12M: {h7_rates['0-12']:.1%}, 49-72M: {h7_rates['49-72']:.1%}",
    '判定': '支持',
    '示唆': '入会後12ヶ月のオンボーディング強化が必要'
})

# H10: Electronic check支払いは離反率が高い
h10_rates = df.groupby('PaymentMethod')['Churn'].apply(lambda x: (x=='Yes').mean())
results.append({
    '仮説': 'H10: Electronic check支払いは離反率が高い',
    '結果': f"E-check: {h10_rates['Electronic check']:.1%}, 他: {h10_rates.drop('Electronic check').mean():.1%}",
    '判定': '支持',
    '示唆': '自動引き落とし/カード決済への移行を促進'
})

report = pd.DataFrame(results)
print(report.to_string(index=False))

import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler, LabelEncoder

df = pd.read_csv('WA_Fn-UseC_-Telco-Customer-Churn.csv')

# 1. 型変換と欠損値処理
df['TotalCharges'] = pd.to_numeric(df['TotalCharges'], errors='coerce').fillna(0)

# 2. customerID削除
df = df.drop('customerID', axis=1)

# 3. 目的変数エンコーディング
df['Churn'] = (df['Churn'] == 'Yes').astype(int)

# 4. カテゴリカル変数のエンコーディング
# 2値変数: LabelEncoding（シンプルで次元が増えない）
binary_cols = ['gender', 'Partner', 'Dependents', 'PhoneService', 'PaperlessBilling']
le = LabelEncoder()
for col in binary_cols:
    df[col] = le.fit_transform(df[col])

# "No internet/phone service" → "No" に統合
service_cols = ['MultipleLines', 'OnlineSecurity', 'OnlineBackup',
                'DeviceProtection', 'TechSupport', 'StreamingTV', 'StreamingMovies']
for col in service_cols:
    df[col] = df[col].replace({'No internet service': 'No', 'No phone service': 'No'})

# 多値変数: One-Hotエンコーディング（順序がないため）
multi_cols = service_cols + ['InternetService', 'Contract', 'PaymentMethod']
df = pd.get_dummies(df, columns=multi_cols, drop_first=True)

# 5. スケーリング（ロジスティック回帰のベースラインモデル用）
numeric_cols = ['tenure', 'MonthlyCharges', 'TotalCharges']
scaler = StandardScaler()

# 6. データ分割 (60:20:20、層化抽出)
X = df.drop('Churn', axis=1)
y = df['Churn']

X_train, X_temp, y_train, y_temp = train_test_split(
    X, y, test_size=0.4, random_state=42, stratify=y)
X_val, X_test, y_val, y_test = train_test_split(
    X_temp, y_temp, test_size=0.5, random_state=42, stratify=y_temp)

# スケーリングは訓練データでfitし、検証・テストにtransform
X_train[numeric_cols] = scaler.fit_transform(X_train[numeric_cols])
X_val[numeric_cols] = scaler.transform(X_val[numeric_cols])
X_test[numeric_cols] = scaler.transform(X_test[numeric_cols])

# 7. 確認
print(f"訓練: {X_train.shape[0]}件 (離反率: {y_train.mean():.3f})")
print(f"検証: {X_val.shape[0]}件 (離反率: {y_val.mean():.3f})")
print(f"テスト: {X_test.shape[0]}件 (離反率: {y_test.mean():.3f})")
print(f"特徴量数: {X_train.shape[1]}")