如何使用 for 循环在决策树上正确实现 bagging？答案

【问题标题】：How to properly implement bagging on decision tree with for loop?如何使用 for 循环在决策树上正确实现 bagging？
【发布时间】：2019-04-12 15:54:25
【问题描述】：

我正在尝试使用决策树和 for 循环来实现 bagging 和投票。我正在使用 sklearn 重采样。但是，我收到了Number of labels=97 does not match number of samples=77，我可以理解为什么，但我不确定如何解决它。

数据集中有 150 个样本。有 150 个标签所以 150 * 0.35 = 97 和 97 * 0.8 = 77。 X 是长度为 150 的特征矩阵，并且 y是长度为150的标签向量

下面是我的代码

from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
from sklearn.utils import resample


X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.35, random_state=3)

predictions = []

for i in range(1,20):
    bootstrap_size = int(0.8*len(X_train))
    bag = resample(X_train, n_samples = bootstrap_size , random_state=i , replace = True) 
    Base_DecisionTree = DecisionTreeClassifier(random_state=3)
    Base_DecisionTree.fit(bag, y_train)
    y_predict = Base_DecisionTree.predict(X_test)
    accuracy = accuracy_score(y_test, y_predict)
    predictions.append(accuracy)

【问题讨论】：

标签： python-3.x machine-learning scikit-learn decision-tree ensemble-learning

【解决方案1】：

您还应该重新采样标签并在fit() 中使用它。

x_bag, y_bag = resample(X_train, y_train, n_samples = bootstrap_size , random_state=i , replace = True) 
tree = DecisionTreeClassifier(random_state=3)
tree.fit(x_bag, y_bag)

【讨论】：

是的，这行得通。但是为什么所有的预测都是一样的呢？ [0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831, 0.8301886792452831]
我很傻，我不得不在 jupyter 上重新启动内核。感谢您的帮助。