Tree Structure

get_params	Extract main parameters of tree ensemble from config.
extract_child_ids	Extract ids of children nodes from table of xgboost model.
transform_variables	Transform variable name in index.
extract_tree_structure	Extract the tree structure.
extract_variable_paths	Extract the variable list of each tree path.
extract_variables	Extract the unique and sorted list of variables from all tree paths.
extract_interactions	Extract the variable interactions of order two from all tree paths.

treehfd.tree_structure.get_params(config: dict) → tuple

Extract main parameters of tree ensemble from config.

treehfd.tree_structure.extract_child_ids(child_id: str) → int

Extract ids of children nodes from table of xgboost model.

treehfd.tree_structure.transform_variables(name: str) → int

Transform variable name in index.

treehfd.tree_structure.extract_tree_structure(tree_table: DataFrame) → tuple

Extract the tree structure.

Parameters:

tree_table (pd.DataFrame) – Dataframe containing the tree structure, derived from the xgboost method trees_to_dataframe.

Returns:

variablesnp.ndarray

array with the variable index for each node split, where -1 indicates terminal leaves.

child_idsnp.ndarray

array with indices of children nodes, with first row for left nodes and second row for right nodes.

split_valuesnp.ndarray

array with the list of node splitting values.

Return type:

tuple

treehfd.tree_structure.extract_variable_paths(tree_structure: tuple, depth_variable: int) → list

Extract the variable list of each tree path.

Parameters:

• tree_structure (tuple) – Tuple containing the splitting variables, children node indices, and splitting node values of the tree.

• depth_variable (int) – Variables are selected at the first depth_variable levels of the tree for the components of the decomposition.

Returns:

List of variable list of each tree path.

Return type:

List

treehfd.tree_structure.extract_variables(variable_paths: list[list[int]]) → ndarray

Extract the unique and sorted list of variables from all tree paths.

Parameters:

variable_paths (list) – List of variable list of each tree path.

Returns:

variable_list – Unique and sorted list of variables of the tree.

Return type:

list

treehfd.tree_structure.extract_interactions(variable_paths: list[list[int]]) → list

Extract the variable interactions of order two from all tree paths.

Parameters:

variable_paths (list) – List of variable list of each tree path.

Returns:

interactions – List of variable index pairs, with a pair for each interaction.

Return type:

list