第8名方案 | 如此简单！

1. 特征工程

• https://www.kaggle.com/code/arunklenin/ps4e4-abalone-age-prediction-regression

2. 训练

• 使用Autogluon进行超参数调优

  hyperparameter_tune_kwargs = {
      'num_trials': 5,
      'scheduler' : 'local',
      'searcher': 'auto',
  }

• 训练设置

  predictor = TabularPredictor(
      label='Rings',
      eval_metric='root_mean_squared_error',
      problem_type='regression',
  )
  predictor.fit(
      train,
      presets='best_quality',
      dynamic_stacking=False,
      num_stack_levels=2,
      excluded_model_types=['KNN', 'NN_TORCH', 'FASTAI'],
      time_limit=3600*24,
      hyperparameter_tune_kwargs=hyperparameter_tune_kwargs,
      keep_only_best=True,
  )

3. 集成

• 参考notebook：https://www.kaggle.com/code/trupologhelper/ps4e4-lightgbm-only
• Autogluon