LightGBM原生接口和Sklearn接口参数详解

• 数据科学：Scipy、Scikit-Learn笔记
• 超参数调优：网格搜索，贝叶斯优化（optuna）详解
• XGBoost原生接口和Sklearn接口参数详解
• LightGBM
• • 一、Sklearn风格接口
  • • lightgbm.LGBMRegressor参数
    • • 1.boosting_type:
      • 2.objective:
      • 3.metric:
      • 4.n_estimators:
      • 5.learning_rate (eta):
      • 6.max_depth:
      • 7.min_child_samples:
      • 8.lambda_l1 (reg_alpha):
      • 9.lambda_l2 (reg_lambda):
      • 10.num_leaves:
      • 11.min_split_gain:
      • 12.early_stopping_rounds:
      • 13.subsample_for_bin:
      • 14.colsample_bytree (feature_fraction):
      • 15.subsample (bagging_fraction):
      • 16.n_jobs:
      • 17.silent (verbose):
    • 调参方法
    • LGBMRegressor.fit参数
    • • 1.X:
      • 2.y:
      • 3.sample_weight:
      • 4.eval_set:
      • 5.eval_names:
      • 6.eval_sample_weight:
      • 7.eval_class_weight:
      • 8.eval_init_score:
      • 9.eval_metric:
      • 10.verbose:
      • 11.callbacks:
      • 12.init_model:
      • 13.pre_partition:
    • LGBMRegressor.predict参数
    • • 1. X
      • 2. num_iteration (n_iter_no_change)
      • 3. raw_score
      • 4. pred_leaf
      • 5. pred_contrib
      • 6. kwargs
  • 二、LightGBM原生接口
  • • 基本设置
    • • 1.boosting_type:
      • 2.objective:
      • 3.metric:
    • 数据处理与采样
    • • 4.max_bin:
      • 5.min_data_in_bin:
      • 6.min_child_samples:
      • 7.subsample:
      • 8.sampling_method:
      • 9.colsample_bytree:
      • 10.colsample_bylevel:
    • 模型训练与正则化
    • • 11.learning_rate:
      • 12.num_leaves:
      • 13.min_split_gain:
      • 14.min_child_weight:
      • 15.lambda_l1/lambda_l2:
      • 16.alpha:
      • 17.n_estimators/num_iterations:
      • 18.early_stopping_rounds:
    • 树结构与剪枝
    • • 19.max_depth:
      • 20.tree_learner:
      • 21.verbosity:
      • 22.gamma:
      • 23.grow_policy:
      • 24.num_threads/n_jobs:
      • 25.distribute: 是否开启分布式训练。默认为 False。
    • 其他
    • • 26.seed:
      • 27.callbacks:
    • lightgbm.train参数
    • • 1.params:
      • 2.train_set:
      • 3.num_boost_round:
      • 4.valid_sets:
      • 5.valid_names:
      • 6.fobj:
      • 7.feval:
      • 8.init_model:
      • 9.feature_name:
      • 10.categorical_feature:
      • 11.early_stopping_rounds:
      • 12.evals_result:
      • 13.verbose_eval:
      • 14.learning_rates:
      • 15.keep_training_booster:
      • 16.callbacks:
      • 17.show_stdv:
    • lightgbm可视化
    • • 1.特征重要性图
      • 2.学习曲线图
      • 3.特征分布直方图
      • 4.特征分裂图
      • 5.决策树结构图
      • 6.特征分布与分裂点图
      • 7.SHAP 贡献值分布图