tabular_trees.BoosterTree

class tabular_trees.BoosterTree(tree, num_leaves, num_cat, split_feature, split_gain, threshold, decision_type, left_child, right_child, leaf_value, leaf_weight, leaf_count, internal_value, internal_weight, internal_count, is_linear, shrinkage)[source]

Bases: object

Data class for individual LightGBM trees.

__init__(tree, num_leaves, num_cat, split_feature, split_gain, threshold, decision_type, left_child, right_child, leaf_value, leaf_weight, leaf_count, internal_value, internal_weight, internal_count, is_linear, shrinkage)

Methods

`__init__`(tree, num_leaves, num_cat, ...)
`get_booster_sting`()	Concatenate tree information to a single string.
`get_booster_string_rows`()	Convert data to rows that could be concatenated to part of BoosterString.

Attributes

`tree`	Tree index.
`num_leaves`	Number of leaves in tree.
`num_cat`
`split_feature`	Split feature indexes for internal nodes.
`split_gain`	Split gain for internal nodes.
`threshold`	Split threshold for internal nodes.
`decision_type`	2 for ordered splits.
`left_child`	Left child node indexes.
`right_child`	Right child node indexes.
`leaf_value`	Leaf predictions.
`leaf_weight`	Sum of Hessian for rows in the leaf node.
`leaf_count`	Number of rows in the leaf node.
`internal_value`	Prediction for internal nodes.
`internal_weight`	Sum of Hessian for rows in the internal node.
`internal_count`	Number of rows in the internal node.
`is_linear`
`shrinkage`

decision_type: 2 for ordered splits.

get_booster_sting()[source]

Concatenate tree information to a single string.

Returns:: model – Booster as a string.
Return type:: str

get_booster_string_rows()[source]: Convert data to rows that could be concatenated to part of BoosterString.

internal_count: Number of rows in the internal node.

internal_value: Prediction for internal nodes.

internal_weight: Sum of Hessian for rows in the internal node.

leaf_count: Number of rows in the leaf node.

leaf_value: Leaf predictions.

leaf_weight: Sum of Hessian for rows in the leaf node.

left_child

Left child node indexes.

Leaf nodes indexes are negative and indexed from -1.

num_leaves: Number of leaves in tree.

right_child

Right child node indexes.

Leaf nodes indexes are negative and indexed from -1.

split_feature: Split feature indexes for internal nodes.

split_gain: Split gain for internal nodes.

threshold: Split threshold for internal nodes.

tree: Tree index.