时间戳

首页 > 解决方案 > 如何在轻 gbm 回归器的贝叶斯优化后查看最佳超参数？

python - 如何在轻 gbm 回归器的贝叶斯优化后查看最佳超参数？

问题描述

我正在使用以下代码，这与我在https://medium.com/analytics-vidhya/hyperparameters-optimization-for-lightgbm-catboost-and-xgboost-regressors-using-bayesian等教程上阅读的内容一致-6e7c495947a9。但是，应该为我的模型打印最佳参数字典的最后一步会导致此错误

''' NameError Traceback (最近一次调用最后一次) in 40 41 bayesion_opt_lgbm(X, y, init_iter=5, n_iters=10, random_state=77, seed = 101, num_iterations = 200) ---> 42 print(optimizer.max )

NameError: name 'optimizer' is not defined ''' 这是我正在使用的完整代码。


from bayes_opt import BayesianOptimization #From the library bayesian-optimization
import warnings
warnings.filterwarnings('ignore')

def bayesion_opt_lgbm(X, y, init_iter=3, n_iters=7, random_state=11, seed = 101, num_iterations = 100):
  dtrain = lgb.Dataset(data=X, label=y)
  def lgb_r2_score(preds, dtrain): #in case want to use R2 as metric
      labels = dtrain.get_label()
      return 'r2', r2_score(labels, preds), True
  # Objective Function
  def hyp_lgbm(num_leaves, feature_fraction, bagging_fraction, max_depth, min_split_gain, min_child_weight):
          params = {'application':'regression','num_iterations': num_iterations,
                    'learning_rate':0.05, 'early_stopping_round':50,
                    'metric':'rmse'} # Default parameters
          params["num_leaves"] = int(round(num_leaves))
          params['feature_fraction'] = max(min(feature_fraction, 1), 0)
          params['bagging_fraction'] = max(min(bagging_fraction, 1), 0)
          params['max_depth'] = int(round(max_depth))
          params['min_split_gain'] = min_split_gain
          params['min_child_weight'] = min_child_weight
          cv_results = lgb.cv(params, dtrain, nfold=5, seed=seed,categorical_feature=[], stratified=False,
                              verbose_eval =None)
          # print(cv_results)
          return np.max(cv_results['rmse-mean'])
  # Domain space-- Range of hyperparameters 
  pds = {'num_leaves': (80, 100),
           'feature_fraction': (0.1, 0.9),
           'bagging_fraction': (0.8, 1),
           'max_depth': (17, 25),
           'min_split_gain': (0.001, 0.1),
           'min_child_weight': (10, 25)
            }

  # Surrogate model
  optimizer = BayesianOptimization(hyp_lgbm, pds, random_state=random_state)
                                    
  # Optimize
  optimizer.maximize(init_points=init_iter, n_iter=n_iters)

bayesion_opt_lgbm(X, y, init_iter=5, n_iters=10, random_state=77, seed = 101, num_iterations = 200)

optimizer.max['params']

为什么'optimizer'在函数中明确定义名称时没有定义名称，我该如何解决？

标签： pythonhyperparameters

解决方案

推荐阅读