scikit-learn decision tree node depth

Question

My goal is to identify at what depth two samples separate within a decision tree. In the development version of scikit-learn you can use the decision_path() method to identify to last common node:

from sklearn import tree
import numpy as np

clf = tree.DecisionTreeClassifier()
clf.fit(data, outcomes)
n_nodes = clf.tree_.node_count
node_indicator = clf.decision_path(data).toarray()
sample_ids = [0,1]
common_nodes = (node_indicator[sample_ids].sum(axis=0) == len(sample_ids))
common_node_id = np.arange(n_nodes)[common_nodes]
max_node = np.max(common_node_id)

Is there a way to determine at what depth the max_node occurs within the tree, possibly with clf.tree_.children_right and clf.tree_.chrildren_left?

maxymoo maxymoo · Accepted Answer · 2016-09-15T01:18:54

Here is a function that you could use to recursively traverse the nodes and calculate the node depths

def get_node_depths(tree):
    """
    Get the node depths of the decision tree

    >>> d = DecisionTreeClassifier()
    >>> d.fit([[1,2,3],[4,5,6],[7,8,9]], [1,2,3])
    >>> get_node_depths(d.tree_)
    array([0, 1, 1, 2, 2])
    """
    def get_node_depths_(current_node, current_depth, l, r, depths):
        depths += [current_depth]
        if l[current_node] != -1 and r[current_node] != -1:
            get_node_depths_(l[current_node], current_depth + 1, l, r, depths)
            get_node_depths_(r[current_node], current_depth + 1, l, r, depths)

    depths = []
    get_node_depths_(0, 0, tree.children_left, tree.children_right, depths) 
    return np.array(depths)

scikit-learn decision tree node depth

1 Answers