AutoML 实战：自动机器学习工具对比与使用

1. AutoML 工具对比

AutoML 工具：
├── Auto-sklearn：基于 scikit-learn，自动选择算法和超参
├── H2O AutoML：企业级，支持分布式
├── TPOT：基于遗传算法的流水线优化
├── AutoKeras：深度学习 AutoML
└── FLAML：微软，轻量快速

2. Auto-sklearn

from autosklearn.classification import AutoSklearnClassifier

automl = AutoSklearnClassifier(
    time_left_for_this_task=300,  # 总时间 5 分钟
    per_run_time_limit=30,        # 单次 30 秒
    n_jobs=-1
)
automl.fit(X_train, y_train)

print(automl.leaderboard())
y_pred = automl.predict(X_test)

3. H2O AutoML

import h2o
from h2o.automl import H2OAutoML

h2o.init()

train = h2o.H2OFrame(pd.concat([X_train, y_train], axis=1))
test = h2o.H2OFrame(X_test)

aml = H2OAutoML(
    max_runtime_secs=300,
    max_models=20,
    seed=42
)
aml.train(y=target_column, training_frame=train)

print(aml.leaderboard())
preds = aml.leader.predict(test)

4. FLAML

from flaml import AutoML

automl = AutoML()
automl.fit(
    X_train, y_train,
    task='classification',
    time_budget=300,
    metric='accuracy',
    estimator_list=['lgbm', 'xgboost', 'rf', 'extra_tree']
)

print(f"最佳模型: {automl.best_estimator}")
print(f"最佳参数: {automl.best_config}")
print(f"最佳分数: {automl.best_loss:.4f}")

总结