Tutorial 30: Capstone — Titanic Survival

import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.compose import ColumnTransformer
from sklearn.pipeline import Pipeline
from sklearn.impute import SimpleImputer
from sklearn.preprocessing import StandardScaler, OneHotEncoder
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import classification_report, accuracy_score

# Load Titanic data
try:
    import seaborn as sns
    df = sns.load_dataset("titanic")
except ImportError:
  raise SystemExit("pip install seaborn")

# Feature engineering
df["family_size"] = df["sibsp"] + df["parch"] + 1
df["is_alone"] = (df["family_size"] == 1).astype(int)
df["deck"] = df["deck"].fillna("Unknown")

features = ["pclass", "sex", "age", "fare", "embarked", "family_size", "is_alone", "deck"]
target = "survived"

data = df[features + [target]].dropna(subset=[target])
X = data[features]
y = data[target]

num_cols = ["age", "fare", "family_size"]
cat_cols = ["pclass", "sex", "embarked", "is_alone", "deck"]

preprocessor = ColumnTransformer([
    ("num", Pipeline([
        ("imputer", SimpleImputer(strategy="median")),
        ("scaler", StandardScaler()),
    ]), num_cols),
    ("cat", Pipeline([
        ("imputer", SimpleImputer(strategy="most_frequent")),
        ("encoder", OneHotEncoder(handle_unknown="ignore")),
    ]), cat_cols),
])

pipe = Pipeline([
    ("prep", preprocessor),
    ("clf", RandomForestClassifier(random_state=42)),
])

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42, stratify=y)

param_grid = {
    "clf__n_estimators": [100, 200],
    "clf__max_depth": [5, 10, None],
    "clf__min_samples_leaf": [1, 3, 5],
}

search = GridSearchCV(pipe, param_grid, cv=5, scoring="accuracy", n_jobs=-1)
search.fit(X_train, y_train)

print(f"Best params: {search.best_params_}")
print(f"CV accuracy: {search.best_score_:.4f}")
y_pred = search.predict(X_test)
print(f"Test accuracy: {accuracy_score(y_test, y_pred):.4f}")
print(classification_report(y_test, y_pred, target_names=["died", "survived"]))

# Feature importances (approximate via permutation on encoded features)
best = search.best_estimator_
importances = best.named_steps["clf"].feature_importances_
print(f"\nTop feature importances (encoded space): {sorted(importances, reverse=True)[:5]}")