XGBoost在设置scale_pos_weight时是否在评估中进行采样

Question

当我为scale_pos_weight赋值时，XGB是否也在评估过程中进行类平衡采样作为训练？

这里有一个设置示例：

Classifier = XGBClassifier(objective="binary:logistic",n_jobs=4,scale_pos_weight=10) 
Classifier.fit(train_features, train_labels,eval_metric="aucpr",eval_set=[(train_features,train_labels),(test_features,test_labels)])

它对我的(test_features,test_labels)进行采样吗？

Answer 1

尝试以下方法：

neg, pos = np.bincount(train_labels)
print(np.bincount(train_labels))
import math 
class_weight1 = math.sqrt(neg/pos) 
print(class_weight1)
class_weight2 = neg/pos
print(class_weight2)
``

分类器=XGB分类器(random_state=0，scale_pos_weight=class_weight 1(

It will resampling the classes in training set only. Resampling techniques are not applicable to test set.