Is it possible to modify / trim learned trees in scikit-learn?

Question

You can access the tree options in sklearn with

tree.tree_.children_left
tree.tree_.children_right
tree.tree_.threshold
tree.tree_.feature

etc.

However, an attempt to write to these variables raises an exception that cannot be written.

Is there a way to modify the learned tree or bypass the AttributeError attribute that cannot be written?

Answer 1

Attributes are like int arrays that cannot be overwritten. You can change the elements of these arrays. This will not facilitate the data.

children_left : array of int, shape [node_count]
    children_left[i] holds the node id of the left child of node i.
    For leaves, children_left[i] == TREE_LEAF. Otherwise,
    children_left[i] > i. This child handles the case where
    X[:, feature[i]] <= threshold[i].

children_right : array of int, shape [node_count]
    children_right[i] holds the node id of the right child of node i.
    For leaves, children_right[i] == TREE_LEAF. Otherwise,
    children_right[i] > i. This child handles the case where
    X[:, feature[i]] > threshold[i].

feature : array of int, shape [node_count]
    feature[i] holds the feature to split on, for the internal node i.

threshold : array of double, shape [node_count]
    threshold[i] holds the threshold for the internal node i.

DecisionTree , . , TREE_LEAF -1.

def prune(decisiontree, min_samples_leaf = 1):
    if decisiontree.min_samples_leaf >= min_samples_leaf:
        raise Exception('Tree already more pruned')
    else:
        decisiontree.min_samples_leaf = min_samples_leaf
        tree = decisiontree.tree_
        for i in range(tree.node_count):
            n_samples = tree.n_node_samples[i]
            if n_samples <= min_samples_leaf:
                tree.children_left[i]=-1
                tree.children_right[i]=-1

, graphviz :

[from sklearn.tree import DecisionTreeRegressor as DTR
from sklearn.datasets import load_diabetes
from sklearn.tree import export_graphviz as export

bunch = load_diabetes()
data = bunch.data
target = bunch.target

dtr = DTR(max_depth = 4)
dtr.fit(data,target)

export(decision_tree=dtr.tree_, out_file='before.dot')
prune(dtr, min_samples_leaf = 100)
export(decision_tree=dtr.tree_, out_file='after.dot')][1]