使用Optuna调参时遇'TypeError: Pipeline对象不可调用'错误求助

Let's break down what's going wrong here and how to fix it quickly.

The Root Cause #

Your error triggers on this line:

optuna_model = xgb_pipeline(**params)

Scikit-Learn's Pipeline isn't a callable function that accepts parameters like XGBClassifier does. You can't instantiate or configure it by calling it with **params — instead, you need to pass your Optuna-tuned hyperparameters to the XGBoost component inside the pipeline, not the pipeline itself.

Solution 1: Use set_params() with Component Prefixes #

Scikit-Learn Pipelines require you to prefix parameters with the component name (followed by two underscores) to target specific steps. Here's how to adjust your code:

def objective(trial):
    params = {
        'max_depth': trial.suggest_int('max_depth', 1, 9),
        'learning_rate': trial.suggest_loguniform('learning_rate', 0.01, 1.0),
        'n_estimators': trial.suggest_int('n_estimators', 50, 500),
        'min_child_weight': trial.suggest_int('min_child_weight', 1, 10),
        'gamma': trial.suggest_loguniform('gamma', 1e-8, 1.0),
        'subsample': trial.suggest_loguniform('subsample', 0.01, 1.0),
        'colsample_bytree': trial.suggest_loguniform('colsample_bytree', 0.01, 1.0),
        'reg_alpha': trial.suggest_loguniform('reg_alpha', 1e-8, 1.0),
        'reg_lambda': trial.suggest_loguniform('reg_lambda', 1e-8, 1.0),
        'eval_metric': 'mlogloss',
        'use_label_encoder': False
    }
    
    # Define base model and pipeline
    xgbmodel = XGBClassifier(random_state=1)
    xgb_pipeline = Pipeline(steps=[
        ('preprocessor', preprocessor),
        ('xgbmodel', xgbmodel)
    ])

    # Add component prefix to params and apply to pipeline
    pipeline_params = {f'xgbmodel__{key}': value for key, value in params.items()}
    xgb_pipeline.set_params(**pipeline_params)

    # Fit and evaluate
    start_time = timer(None)
    xgb_pipeline.fit(X_train, y_train)
    y_pred = xgb_pipeline.predict(X_valid)
    accuracy = accuracy_score(y_valid, y_pred)
    return accuracy

Solution 2: Pass Params Directly to XGBClassifier #

A simpler alternative is to initialize your XGBClassifier with the Optuna params first, then add it to the pipeline:

def objective(trial):
    params = {
        'max_depth': trial.suggest_int('max_depth', 1, 9),
        'learning_rate': trial.suggest_loguniform('learning_rate', 0.01, 1.0),
        'n_estimators': trial.suggest_int('n_estimators', 50, 500),
        'min_child_weight': trial.suggest_int('min_child_weight', 1, 10),
        'gamma': trial.suggest_loguniform('gamma', 1e-8, 1.0),
        'subsample': trial.suggest_loguniform('subsample', 0.01, 1.0),
        'colsample_bytree': trial.suggest_loguniform('colsample_bytree', 0.01, 1.0),
        'reg_alpha': trial.suggest_loguniform('reg_alpha', 1e-8, 1.0),
        'reg_lambda': trial.suggest_loguniform('reg_lambda', 1e-8, 1.0),
        'eval_metric': 'mlogloss',
        'use_label_encoder': False,
        'random_state': 1  # Move random state into params
    }
    
    # Initialize model with trial params
    xgbmodel = XGBClassifier(**params)
    
    # Build pipeline with preconfigured model
    xgb_pipeline = Pipeline(steps=[
        ('preprocessor', preprocessor),
        ('xgbmodel', xgbmodel)
    ])

    # Fit and evaluate
    start_time = timer(None)
    xgb_pipeline.fit(X_train, y_train)
    y_pred = xgb_pipeline.predict(X_valid)
    accuracy = accuracy_score(y_valid, y_pred)
    return accuracy

Quick Note #

If you're using a newer version of XGBoost, use_label_encoder is deprecated — you can remove that parameter entirely since it's no longer needed.

内容的提问来源于stack exchange，提问作者Steve Rowe