🎛️ 05 — Model Selection and Tuning

Model selection compares candidate models using fair evaluation.

---

Grid Search

from sklearn.model_selection import GridSearchCV

param_grid = {
    "model__n_estimators": [100, 200],
    "model__max_depth": [None, 5, 10]
}

search = GridSearchCV(
    model,
    param_grid,
    cv=5,
    scoring="accuracy"
)

search.fit(X_train, y_train)
print(search.best_params_)

---

Random Search

from sklearn.model_selection import RandomizedSearchCV

Random search is often faster for large parameter spaces.

---

Rules

• tune on training/cross-validation
• evaluate final model once on test set
• keep metric aligned with business goal
• avoid chasing tiny leaderboard gains

---

Next

➡️ 06-exercises