🏡 Housing EBM + DoWhy 리포트

from sklearn.datasets import fetch_california_housing
import pandas as pd

data = fetch_california_housing(as_frame=True)
housing = data.frame  # 20,640 rows

# IQR 이상치 제거 (결과 변수 제외)
outlier_cols = ['AveRooms', 'AveOccup', 'Population']
mask = pd.Series(True, index=housing.index)
for col in outlier_cols:
    Q1, Q3 = housing[col].quantile([0.25, 0.75])
    IQR = Q3 - Q1
    mask &= (housing[col] >= Q1 - 1.5*IQR) & (housing[col] <= Q3 + 1.5*IQR)
housing_clean = housing[mask]  # 18,350 rows

from interpret.glassbox import ExplainableBoostingRegressor
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error, r2_score
import numpy as np

X = housing_clean.drop(columns=['MedHouseVal'])
y = housing_clean['MedHouseVal']

X_temp, X_test, y_temp, y_test = train_test_split(
    X, y, test_size=0.1, random_state=42)
X_train, X_val, y_train, y_val = train_test_split(
    X_temp, y_temp, test_size=0.1111, random_state=42)

ebm = ExplainableBoostingRegressor(
    max_bins=256,
    interactions=5,
    random_state=42
)
ebm.fit(X_train, y_train)

val_pred = ebm.predict(X_val)
print(f"Val RMSE: {np.sqrt(mean_squared_error(y_val, val_pred)):.4f}")
print(f"Val R²:   {r2_score(y_val, val_pred):.4f}")
print(f"Test R²:  {r2_score(y_test, ebm.predict(X_test)):.4f}")

from dowhy import CausalModel

model = CausalModel(
    data=housing_clean,
    treatment='MedInc',
    outcome='MedHouseVal',
    graph="""
        digraph {
            Latitude -> MedInc; Latitude -> MedHouseVal;
            Longitude -> MedInc; Longitude -> MedHouseVal;
            AveRooms -> MedInc; AveRooms -> MedHouseVal;
            AveOccup -> MedInc; AveOccup -> MedHouseVal;
            MedInc -> MedHouseVal;
        }
    """
)

identified = model.identify_effect(proceed_when_unidentifiable=True)
estimate = model.estimate_effect(
    identified,
    method_name="backdoor.linear_regression"
)
print(f"ATE: {estimate.value:.4f}")  # 단위: $100,000