Optunaで学ぶハイパーパラメーターチューニング — 仕組みと実践ガイド

import optuna
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import cross_val_score

def objective(trial):
    """Optuna が最小化/最大化する目的関数。"""
    # パラメーター空間を動的に定義
    n_estimators = trial.suggest_int("n_estimators", 50, 500)
    max_depth = trial.suggest_int("max_depth", 3, 20)
    min_samples_split = trial.suggest_int("min_samples_split", 2, 20)
    min_samples_leaf = trial.suggest_int("min_samples_leaf", 1, 10)

    clf = RandomForestClassifier(
        n_estimators=n_estimators,
        max_depth=max_depth,
        min_samples_split=min_samples_split,
        min_samples_leaf=min_samples_leaf,
        random_state=42,
    )
    # 5-fold CV の平均精度を返す
    score = cross_val_score(clf, X_train, y_train, cv=5).mean()
    return score

# Study を作成して最適化を実行
study = optuna.create_study(direction="maximize")
study.optimize(objective, n_trials=100)

print(f"Best score: {study.best_value:.4f}")
print(f"Best params: {study.best_params}")

import optuna
from sklearn.model_selection import cross_val_score
from xgboost import XGBClassifier

def objective(trial):
    params = {
        "n_estimators": trial.suggest_int("n_estimators", 100, 1000),
        "max_depth": trial.suggest_int("max_depth", 3, 12),
        "learning_rate": trial.suggest_float("learning_rate", 1e-3, 0.3, log=True),
        "subsample": trial.suggest_float("subsample", 0.5, 1.0),
        "colsample_bytree": trial.suggest_float("colsample_bytree", 0.5, 1.0),
        "reg_alpha": trial.suggest_float("reg_alpha", 1e-8, 10.0, log=True),
        "reg_lambda": trial.suggest_float("reg_lambda", 1e-8, 10.0, log=True),
    }

    clf = XGBClassifier(**params, random_state=42, use_label_encoder=False)

    # Pruning: 各 fold の途中結果を報告
    scores = []
    for fold_idx, (train_idx, val_idx) in enumerate(kf.split(X, y)):
        clf.fit(X[train_idx], y[train_idx])
        score = clf.score(X[val_idx], y[val_idx])
        scores.append(score)

        # 途中経過を報告 → 見込みが低ければ自動打ち切り
        trial.report(sum(scores) / len(scores), fold_idx)
        if trial.should_prune():
            raise optuna.TrialPruned()

    return sum(scores) / len(scores)

study = optuna.create_study(
    direction="maximize",
    pruner=optuna.pruners.MedianPruner(n_startup_trials=10),
)
study.optimize(objective, n_trials=200)

def objective(trial):
    # アルゴリズム自体も探索対象にできる
    algo = trial.suggest_categorical("algorithm", ["rf", "xgb", "lgbm"])

    if algo == "rf":
        params = {
            "n_estimators": trial.suggest_int("rf_n_estimators", 50, 500),
            "max_depth": trial.suggest_int("rf_max_depth", 3, 20),
        }
        model = RandomForestClassifier(**params)

    elif algo == "xgb":
        params = {
            "n_estimators": trial.suggest_int("xgb_n_estimators", 100, 1000),
            "learning_rate": trial.suggest_float("xgb_lr", 1e-3, 0.3, log=True),
            "max_depth": trial.suggest_int("xgb_max_depth", 3, 12),
        }
        model = XGBClassifier(**params)

    else:  # lgbm
        params = {
            "n_estimators": trial.suggest_int("lgbm_n_estimators", 100, 1000),
            "learning_rate": trial.suggest_float("lgbm_lr", 1e-3, 0.3, log=True),
            "num_leaves": trial.suggest_int("lgbm_num_leaves", 15, 127),
        }
        model = LGBMClassifier(**params)

    return cross_val_score(model, X, y, cv=5).mean()

import optuna.visualization as vis

# 最適化の履歴（各トライアルのスコア推移）
vis.plot_optimization_history(study)

# パラメーターの重要度（どのパラメーターがスコアに最も影響するか）
vis.plot_param_importances(study)

# パラメーター間の関係をヒートマップで表示
vis.plot_contour(study, params=["learning_rate", "max_depth"])

# 各パラメーターとスコアの関係をスライスプロットで確認
vis.plot_slice(study)

# 並行座標プロット（全パラメーターの組み合わせを俯瞰）
vis.plot_parallel_coordinate(study)