gbdt模型训练遇到报错

Terence888

代码如下

#Gradient Boost Decision Tree GBDT
from sklearn.ensemble import GradientBoostingClassifier
from sklearn.model_selection import GridSearchCV
from sklearn.metrics import accuracy_score

gbc = GradientBoostingClassifier()
param_test = {'n_estimators': [50, 100], 
              'min_samples_split': [3, 4, 5, 6, 7],
              'max_depth': [3, 4, 5, 6]}
gbc_grid = GridSearchCV(gbc, param_test, cv=5, refit=True, verbose=1)
gbc_grid.fit(X_train,y_train)
train_score = gbc.score(X_train,y_train)
gbc_pre = gbc.predict(X_valid)
valid_score = accuracy_score(y_valid,gbc_pre)
print(valid_score)

报错如下

Fitting 5 folds for each of 40 candidates, totalling 200 fits
/opt/conda/lib/python3.10/site-packages/sklearn/base.py:432: UserWarning: X has feature names, but GradientBoostingClassifier was fitted without feature names
  warnings.warn(
---------------------------------------------------------------------------
NotFittedError                            Traceback (most recent call last)
Cell In[25], line 12
     10 gbc_grid = GridSearchCV(gbc, param_test, cv=5, refit=True, verbose=1)
     11 gbc_grid.fit(X_train,y_train)
---> 12 train_score = gbc.score(X_train,y_train)
     13 gbc_pre = gbc.predict(X_valid)
     14 valid_score = accuracy_score(y_valid,gbc_pre)

File /opt/conda/lib/python3.10/site-packages/sklearn/base.py:668, in ClassifierMixin.score(self, X, y, sample_weight)
    643 """
    644 Return the mean accuracy on the given test data and labels.
    645 
   (...)
    664     Mean accuracy of ``self.predict(X)`` w.r.t. `y`.
    665 """
    666 from .metrics import accuracy_score
--> 668 return accuracy_score(y, self.predict(X), sample_weight=sample_weight)

File /opt/conda/lib/python3.10/site-packages/sklearn/ensemble/_gb.py:1308, in GradientBoostingClassifier.predict(self, X)
   1293 def predict(self, X):
   1294     """Predict class for X.
   1295 
   1296     Parameters
   (...)
   1306         The predicted values.
   1307     """
-> 1308     raw_predictions = self.decision_function(X)
   1309     encoded_labels = self._loss._raw_prediction_to_decision(raw_predictions)
   1310     return self.classes_.take(encoded_labels, axis=0)

File /opt/conda/lib/python3.10/site-packages/sklearn/ensemble/_gb.py:1264, in GradientBoostingClassifier.decision_function(self, X)
   1243 """Compute the decision function of ``X``.
   1244 
   1245 Parameters
   (...)
   1259     array of shape (n_samples,).
   1260 """
   1261 X = self._validate_data(
   1262     X, dtype=DTYPE, order="C", accept_sparse="csr", reset=False
   1263 )
-> 1264 raw_predictions = self._raw_predict(X)
   1265 if raw_predictions.shape[1] == 1:
   1266     return raw_predictions.ravel()

File /opt/conda/lib/python3.10/site-packages/sklearn/ensemble/_gb.py:687, in BaseGradientBoosting._raw_predict(self, X)
    685 def _raw_predict(self, X):
    686     """Return the sum of the trees raw predictions (+ init estimator)."""
--> 687     raw_predictions = self._raw_predict_init(X)
    688     predict_stages(self.estimators_, X, self.learning_rate, raw_predictions)
    689     return raw_predictions

File /opt/conda/lib/python3.10/site-packages/sklearn/ensemble/_gb.py:673, in BaseGradientBoosting._raw_predict_init(self, X)
    671 def _raw_predict_init(self, X):
    672     """Check input and compute raw predictions of the init estimator."""
--> 673     self._check_initialized()
    674     X = self.estimators_[0, 0]._validate_X_predict(X, check_input=True)
    675     if self.init_ == "zero":

File /opt/conda/lib/python3.10/site-packages/sklearn/ensemble/_gb.py:382, in BaseGradientBoosting._check_initialized(self)
    380 def _check_initialized(self):
    381     """Check that the estimator is initialized, raising an error if not."""
--> 382     check_is_fitted(self)

File /opt/conda/lib/python3.10/site-packages/sklearn/utils/validation.py:1390, in check_is_fitted(estimator, attributes, msg, all_or_any)
   1385     fitted = [
   1386         v for v in vars(estimator) if v.endswith("_") and not v.startswith("__")
   1387     ]
   1389 if not fitted:
-> 1390     raise NotFittedError(msg % {"name": type(estimator).__name__})

NotFittedError: This GradientBoostingClassifier instance is not fitted yet. Call 'fit' with appropriate arguments before using this estimator.

FishC

从报错信息来看，`GradientBoostingClassifier`实例`gbc`没有经过训练，所以无法进行评分或者预测。这是因为你在`GridSearchCV`中训练的是`gbc_grid`，而你试图用`gbc`进行评分和预测。

以下是你需要做的更改：

1. 使用`gbc_grid.best_estimator_`来获取经过训练的最佳模型实例。

2. 将`gbc.score(X_train, y_train)`改为`gbc_grid.score(X_train, y_train)`。

3. 将`gbc.predict(X_valid)`改为`gbc_grid.predict(X_valid)`。

修改后的代码应该如下所示：

param_test = {'n_estimators': [100, 200], 
              'min_samples_split': [2, 5],
              'max_depth': [3, 4]}
gbc_grid = GridSearchCV(gbc, param_test, cv=5, refit=True, verbose=1)

# Train the model using X_train and y_train
gbc_grid.fit(X_train, y_train)

# Use the best estimator found
train_score = gbc_grid.score(X_train, y_train)
gbc_pre = gbc_grid.predict(X_valid)
valid_score = accuracy_score(y_valid, gbc_pre)

print(valid_score)

请注意，`param_test`中的参数列表应该包含具体的值供GridSearchCV搜索。另外，确保你传入的`X_train`和`X_valid`数据形式与模型的期望一致。

以上回复来自 FishC AI 小助理，如未能正确解答您的问题，请继续追问。