Scikit-learn 决策树提取特征节点

Question

我有一个决策树，它在包含七个特征的数据集上进行了训练。我想要一些方法来轻松提取每个节点，其中数据在特定特征上被分割，并且哪个阈值或类别被用作分割标准。这是我用来生成树的小模板。clfSCReduced = DecisionTreeClassifier() clfSCReduced.fit(featuresSC, labelsSC)有什么建议吗？

Answer 1

从DecisionTreeClassifier 文档中，它声明您可以从中获取Tree对象clsfSCReduced.tree_。scikit-learn 文档在这里有一个关于如何从树中获取信息的示例。该示例提供以下输出：

The binary tree structure has 5 nodes and has the following tree structure:

node=0 test node: go to node 1 if X[:, 3] <= 0.800000011920929 else to node 2.

        node=1 leaf node.

        node=2 test node: go to node 3 if X[:, 2] <= 4.950000047683716 else to node 4.

                node=3 leaf node.

                node=4 leaf node.

Rules used to predict sample 0:

decision id node 0 : (X_test[0, 3] (= 2.4) > 0.800000011920929)

decision id node 2 : (X_test[0, 2] (= 5.1) > 4.950000047683716)

The following samples [0, 1] share the node [0 2] in the tree

It is 40.0 % of all nodes.

我在下面复制了他们的示例以完成

import numpy as np

from sklearn.model_selection import train_test_split

from sklearn.datasets import load_iris

from sklearn.tree import DecisionTreeClassifier

iris = load_iris()

X = iris.data

y = iris.target

X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)

estimator = DecisionTreeClassifier(max_leaf_nodes=3, random_state=0)

estimator.fit(X_train, y_train)

# The decision estimator has an attribute called tree_  which stores the entire

# tree structure and allows access to low level attributes. The binary tree

# tree_ is represented as a number of parallel arrays. The i-th element of each

# array holds information about the node `i`. Node 0 is the tree's root. NOTE:

# Some of the arrays only apply to either leaves or split nodes, resp. In this

# case the values of nodes of the other type are arbitrary!

#

# Among those arrays, we have:

#   - left_child, id of the left child of the node

#   - right_child, id of the right child of the node

#   - feature, feature used for splitting the node

#   - threshold, threshold value at the node

#

# Using those arrays, we can parse the tree structure:

n_nodes = estimator.tree_.node_count

children_left = estimator.tree_.children_left

children_right = estimator.tree_.children_right

feature = estimator.tree_.feature

threshold = estimator.tree_.threshold

# The tree structure can be traversed to compute various properties such

# as the depth of each node and whether or not it is a leaf.

node_depth = np.zeros(shape=n_nodes, dtype=np.int64)

is_leaves = np.zeros(shape=n_nodes, dtype=bool)

stack = [(0, -1)]  # seed is the root node id and its parent depth

while len(stack) > 0:

    node_id, parent_depth = stack.pop()

    node_depth[node_id] = parent_depth + 1

    # If we have a test node

    if (children_left[node_id] != children_right[node_id]):

        stack.append((children_left[node_id], parent_depth + 1))

        stack.append((children_right[node_id], parent_depth + 1))

    else:

        is_leaves[node_id] = True

print("The binary tree structure has %s nodes and has "

      "the following tree structure:"

      % n_nodes)

for i in range(n_nodes):

    if is_leaves[i]:

        print("%snode=%s leaf node." % (node_depth[i] * "\t", i))

    else:

        print("%snode=%s test node: go to node %s if X[:, %s] <= %s else to "

              "node %s."

              % (node_depth[i] * "\t",

                 i,

                 children_left[i],

                 feature[i],

                 threshold[i],

                 children_right[i],

                 ))

print()

# First let's retrieve the decision path of each sample. The decision_path

# method allows to retrieve the node indicator functions. A non zero element of

# indicator matrix at the position (i, j) indicates that the sample i goes

# through the node j.

node_indicator = estimator.decision_path(X_test)

# Similarly, we can also have the leaves ids reached by each sample.

leave_id = estimator.apply(X_test)

# Now, it's possible to get the tests that were used to predict a sample or

# a group of samples. First, let's make it for the sample.

sample_id = 0

node_index = node_indicator.indices[node_indicator.indptr[sample_id]:

                                    node_indicator.indptr[sample_id + 1]]

print('Rules used to predict sample %s: ' % sample_id)

for node_id in node_index:

    if leave_id[sample_id] == node_id:

        continue

    if (X_test[sample_id, feature[node_id]] <= threshold[node_id]):

        threshold_sign = "<="

    else:

        threshold_sign = ">"

    print("decision id node %s : (X_test[%s, %s] (= %s) %s %s)"

          % (node_id,

             sample_id,

             feature[node_id],

             X_test[sample_id, feature[node_id]],

             threshold_sign,

             threshold[node_id]))

# For a group of samples, we have the following common node.

sample_ids = [0, 1]

common_nodes = (node_indicator.toarray()[sample_ids].sum(axis=0) ==

                len(sample_ids))

common_node_id = np.arange(n_nodes)[common_nodes]

print("\nThe following samples %s share the node %s in the tree"

      % (sample_ids, common_node_id))

print("It is %s %% of all nodes." % (100 * len(common_node_id) / n_nodes,))

Answer 2

我也在这里提供主要思想。以下代码来自 sklearn 文档，并进行了一些小改动以实现您的目标。

import numpy as np

from sklearn.model_selection import train_test_split

from sklearn.datasets import load_iris

from sklearn.tree import DecisionTreeClassifier

iris = load_iris()

X = iris.data

y = iris.target

X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)

estimator = DecisionTreeClassifier(max_leaf_nodes=3, random_state=0)

estimator.fit(X_train, y_train)

# The decision estimator has an attribute called tree_  which stores the entire

# tree structure and allows access to low level attributes. The binary tree

# tree_ is represented as a number of parallel arrays. The i-th element of each

# array holds information about the node `i`. Node 0 is the tree's root. NOTE:

# Some of the arrays only apply to either leaves or split nodes, resp. In this

# case the values of nodes of the other type are arbitrary!

#

# Among those arrays, we have:

#   - left_child, id of the left child of the node

#   - right_child, id of the right child of the node

#   - feature, feature used for splitting the node

#   - threshold, threshold value at the node

n_nodes = estimator.tree_.node_count

children_left = estimator.tree_.children_left

children_right = estimator.tree_.children_right

feature = estimator.tree_.feature

threshold = estimator.tree_.threshold

# The tree structure can be traversed to compute various properties such

# as the depth of each node and whether or not it is a leaf.

node_depth = np.zeros(shape=n_nodes, dtype=np.int64)

is_leaves = np.zeros(shape=n_nodes, dtype=bool)

stack = [(0, -1)]  # seed is the root node id and its parent depth

while len(stack) > 0:

    node_id, parent_depth = stack.pop()

    node_depth[node_id] = parent_depth + 1

    # If we have a test node

    if (children_left[node_id] != children_right[node_id]):

        stack.append((children_left[node_id], parent_depth + 1))

        stack.append((children_right[node_id], parent_depth + 1))

    else:

        is_leaves[node_id] = True

print("The binary tree structure has %s nodes and has "

      "the following tree structure:"

      % n_nodes)

for i in range(n_nodes):

    if is_leaves[i]:

        print("%snode=%s leaf node." % (node_depth[i] * "\t", i))

    else:

        print("%snode=%s test node: go to node %s if X[:, %s] <= %s else to "

              "node %s."

              % (node_depth[i] * "\t",

                 i,

                 children_left[i],

                 feature[i],

                 threshold[i],

                 children_right[i],

                 ))

print("\n")

# First let's retrieve the decision path of each sample. The decision_path

# method allows to retrieve the node indicator functions. A non zero element of

# indicator matrix at the position (i, j) indicates that the sample i goes

# through the node j.

node_indicator = estimator.decision_path(X_test)

# Similarly, we can also have the leaves ids reached by each sample.

leave_id = estimator.apply(X_test)

# Now, it's possible to get the tests that were used to predict a sample or

# a group of samples. First, let's make it for the sample.

# HERE IS WHAT YOU WANT

sample_id = 0

node_index = node_indicator.indices[node_indicator.indptr[sample_id]:

                                    node_indicator.indptr[sample_id + 1]]

print('Rules used to predict sample %s: ' % sample_id)

for node_id in node_index:

    if leave_id[sample_id] == node_id:  # <-- changed != to ==

        #continue # <-- comment out

        print("leaf node {} reached, no decision here".format(leave_id[sample_id])) # <--

    else: # < -- added else to iterate through decision nodes

        if (X_test[sample_id, feature[node_id]] <= threshold[node_id]):

            threshold_sign = "<="

        else:

            threshold_sign = ">"

        print("decision id node %s : (X[%s, %s] (= %s) %s %s)"

              % (node_id,

                 sample_id,

                 feature[node_id],

                 X_test[sample_id, feature[node_id]], # <-- changed i to sample_id

                 threshold_sign,

                 threshold[node_id]))

这将在最后打印以下内容：

Rules used to predict sample 0: 

decision id node 0 : (X[0, 3] (= 2.4) > 0.800000011920929)

decision id node 2 : (X[0, 2] (= 5.1) > 4.950000047683716)

leaf node 4 reached, no decision here