from sklearn.model_selection import cross_val_score, GridSearchCV
from sklearn.ensemble import RandomForestClassifier
from sklearn.preprocessing import StandardScaler

# Feature engineering
X = df[features]
y = df['target']

scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)

# Model selection
param_grid = {
    'n_estimators': [100, 200, 300],
    'max_depth': [10, 20, None],
    'min_samples_split': [2, 5, 10]
}

model = RandomForestClassifier(random_state=42)
grid_search = GridSearchCV(model, param_grid, cv=5, scoring='f1')
grid_search.fit(X_scaled, y)

print("[FINDING]")
print(f"Best model: {grid_search.best_params_}")

print("[STAT:CV_SCORE]")
scores = cross_val_score(grid_search.best_estimator_, X_scaled, y, cv=5)
print(f"CV F1: {scores.mean():.4f} ± {scores.std():.4f}")

print("[LIMITATION]")
print("Model assumes feature independence. Consider feature selection.")

from statsmodels.stats.proportion import proportions_ztest

# A/B test analysis
conversions = [treatment_conversions, control_conversions]
totals = [treatment_total, control_total]

z_stat, p_value = proportions_ztest(conversions, totals)
lift = (conversions[0]/totals[0] - conversions[1]/totals[1]) / (conversions[1]/totals[1])

print("[FINDING]")
print(f"Treatment shows {lift*100:.1f}% lift over control")

print("[STAT:SIGNIFICANCE]")
print(f"Z-score: {z_stat:.2f}, p-value: {p_value:.4f}")