Tree Utilities

Tree parsing utilities and lightweight tree data structures.

class Node(node_id, *, feature=None, value=None, parent=None, threshold=None, code=None, n_samples=0, left=None, right=None)[source]

Bases: NodeMixin

Node in a parsed decision tree.

property code

property feature

property left

property length

property n_samples

property node_id

property right

property sigma

property threshold

property value

class Tree(root)[source]

Bases: object

In-memory representation of a parsed tree ensemble member.

property adaboost

property leaves

property logit

property max_depth

property n_nodes

nodes_at(depth)[source]

property shape

property xgboost

root

parse_ensembles(*ensembles, mapper)[source]

Flatten and parse one or more supported tree ensembles.

Returns:: Parsed trees from every provided ensemble, flattened into one tuple.
Return type:: tuple[Tree, …]

parse_tree(tree, *, mapper, is_adaboost=False)[source]

Convert a fitted scikit-learn tree into an OCEAN tree.

Returns:: Parsed tree structure used by the optimization backends.
Return type:: Tree

parse_trees(trees, *, mapper, is_adaboost=False)[source]

Parse an iterable of fitted scikit-learn trees.

Returns:: Parsed tree structures in the same order as the input iterable.
Return type:: tuple[Tree, …]