我正在使用python 3.4中的scikit-learn包中的决策树分类器,并且我想为每个输入数据点获得相应的叶节点id。
例如,我的输入可能像这样:
array([[ 5.1, 3.5, 1.4, 0.2],
[ 4.9, 3. , 1.4, 0.2],
[ 4.7, 3.2, 1.3, 0.2]])
,假设对应的叶节点分别为16、5和45。我希望我的输出是:
leaf_node_id = array([16, 5, 45])
我已经阅读了scikit-learn邮件列表和SF的相关问题,但我仍然不能使它工作。这是我在邮件列表中找到的一些提示,但仍然不起作用。
http://sourceforge.net/p/scikit-learn/mailman/message/31728624/在一天结束时,我只想有一个函数GetLeafNode(clf, x_validate),这样它的输出是相应叶节点的列表。下面是再现我收到的错误的代码。所以,任何建议都将是非常感谢的。
from sklearn.datasets import load_iris
from sklearn import tree
# load data and divide it to train and validation
iris = load_iris()
num_train = 100
X_train = iris.data[:num_train,:]
X_valida = iris.data[num_train:,:]
y_train = iris.target[:num_train]
y_valida = iris.target[num_train:]
# fit the decision tree using the train data set
clf = tree.DecisionTreeClassifier()
clf = clf.fit(X_train, y_train)
# Now I want to know the corresponding leaf node id for each of my training data point
clf.tree_.apply(X_train)
# This gives the error message below:
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-17-2ecc95213752> in <module>()
----> 1 clf.tree_.apply(X_train)
_tree.pyx in sklearn.tree._tree.Tree.apply (sklearn/tree/_tree.c:19595)()
ValueError: Buffer dtype mismatch, expected 'DTYPE_t' but got 'double'
自scikit-learn 0.17以来,您可以使用DecisionTree对象的apply方法来获取数据点在树中结束的叶子的索引。基于neobot的回答:
from sklearn.datasets import load_iris
from sklearn import tree
# load data and divide it to train and validation
iris = load_iris()
num_train = 100
X_train = iris.data[:num_train,:]
X_valida = iris.data[num_train:,:]
y_train = iris.target[:num_train]
y_valida = iris.target[num_train:]
# fit the decision tree using the train data set
clf = tree.DecisionTreeClassifier()
clf = clf.fit(X_train, y_train)
# Compute the leaf node id for each of my training data points
clf.apply(X_train)
产生输出
array([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
2, 2, 2, 2, 2, 2, 2, 2])
我终于让它工作了。以下是一个基于我在scikit-learn邮件列表中的通信信息的解决方案:
在scikit-learn 0.16.1版本之后,在clf.tree_
中实现了apply方法,因此,我遵循了以下步骤:
- 将scikit-learn更新到最新版本(0.16.1),以便您可以从
clf.tree_
使用 - 使用
X_train = X_train.astype('float32')
将输入数据数组( - 现在你可以这样使用
apply
方法:clf.tree_.apply(X_train)
,你会得到每个数据点的叶子节点id。
apply
方法X_train
, X_valida
)从float64
转换为float32
from sklearn.datasets import load_iris
from sklearn import tree
# load data and divide it to train and validation
iris = load_iris()
num_train = 100
X_train = iris.data[:num_train,:]
X_valida = iris.data[num_train:,:]
y_train = iris.target[:num_train]
y_valida = iris.target[num_train:]
# convert data to float32
X_train = X_train.astype('float32')
# fit the decision tree using the train data set
clf = tree.DecisionTreeClassifier()
clf = clf.fit(X_train, y_train)
# Now I want to know the corresponding leaf node id for each of my training data point
clf.tree_.apply(X_train)
# This gives the leaf node id:
array([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
2, 2, 2, 2, 2, 2, 2, 2])