CatboostRegressor可以接收哪种类型的数据集？

Question

我一直在尝试一个我感兴趣的代码项目。我的数据集来自外汇市场，有10个特征，超过70000个属性，已经被分成了一个训练和测试集，但我的CatboostRegressor一直给出这个错误。我需要对我的数据集执行什么操作才能使回归器工作？或者是别的什么？？

from catboost import CatBoostRegressor, Pool
train_data = Pool(zar_train, label=['bidclose', 'askclose'])

test_data = Pool(zar_test, label=['bidclose', 'askclose'])
eval_data = zar_val
eval_dataset = Pool(eval_data, label=['bidclose', 'askclose'])

model = CatBoostRegressor(learning_rate=0.03,
                           custom_metric=['Logloss',
                                          'AUC:hints=skip_train~false'], score_function='Accuracy')

model.fit(train_data, test_data)

print(model.get_best_score())

错误：

---------------------------------------------------------------------------
CatBoostError                             Traceback (most recent call last)
<ipython-input-8-e0aa9e711bf9> in <module>
      1 from catboost import CatBoostRegressor, Pool
----> 2 train_data = Pool(zar_train, label=['bidclose', 'askclose'])
      3 test_data = Pool(zar_test, label=['bidclose', 'askclose'])
      4 eval_data = zar_val
      5 eval_dataset = Pool(eval_data, label=['bidclose', 'askclose'])

~\anaconda3\lib\site-packages\catboost\core.py in __init__(self, data, label, cat_features, text_features, embedding_features, column_description, pairs, delimiter, has_header, ignore_csv_quoting, weight, group_id, group_weight, subgroup_id, pairs_weight, baseline, feature_names, thread_count, log_cout, log_cerr)
    615                     )
    616 
--> 617                 self._init(data, label, cat_features, text_features, embedding_features, pairs, weight, group_id, group_weight, subgroup_id, pairs_weight, baseline, feature_names, thread_count)
    618         super(Pool, self).__init__()
    619 

~\anaconda3\lib\site-packages\catboost\core.py in _init(self, data, label, cat_features, text_features, embedding_features, pairs, weight, group_id, group_weight, subgroup_id, pairs_weight, baseline, feature_names, thread_count)
   1085             if len(np.shape(label)) == 1:
   1086                 label = np.expand_dims(label, 1)
-> 1087             self._check_label_shape(label, samples_count)
   1088         if feature_names is not None:
   1089             self._check_feature_names(feature_names, features_count)

~\anaconda3\lib\site-packages\catboost\core.py in _check_label_shape(self, label, samples_count)
    730         """
    731         if len(label) != samples_count:
--> 732             raise CatBoostError("Length of label={} and length of data={} is different.".format(len(label), samples_count))
    733 
    734     def _check_baseline_type(self, baseline):

CatBoostError: Length of label=2 and length of data=44908 is different.

Answer 1

文档中说data必须是2D数组，label必须是具有相同长度的的一维数组。请参阅here

由于您使用了包含2个值的列表作为label，因此出现错误。

相反，您应该传递整个y_train和y_test值，它们应该等于数据集的长度。

train_data = Pool(zar_train, label=y_train) #length of zar_train and y_train should be equal

test_dataset = Pool(zar_test, label=y_test) #length of test_data and y_test should be equal

eval_dataset = Pool(eval_data, label=y_eval) #length of eval_data and y_eval should be equal