api documentation

Trees

export_tree_data()	Export tree data from model.
TabularTrees(trees, get_root_node_given_tree)	Generic tree structure in tabular format.

LightGBM

LightGBMTabularTrees(data)	Class to hold the LightGBM trees in tabular format.
EditableBooster(header, trees, bottom_rows)	Editable LightGBM booster.
BoosterHeader(header, version, num_class, ...)	Dataclass for the metadata in header section of a Booster.
BoosterTree(tree, num_leaves, num_cat, ...)	Data class for individual LightGBM trees.
BoosterString(rows)	String version of a LightGBM Booster.

Scikit-Learn

ScikitLearnTabularTrees(data)	Scikit-Learn GradientBoosting trees in tabular format.
ScikitLearnHistTabularTrees(data)	Scikit-Learn HistGradientBoosting trees in tabular format.

XGBoost

XGBoostTabularTrees(data)	Class to hold the XGBoost trees in tabular format.
XGBoostParser(model[, reader])	Class that dumps an xgboost Booster then parses the dumped file.
ParsedXGBoostTabularTrees(data)	Dataclass for XGBoost models that have been parsed from a model dump.

Warning

The XGBoostParser is depreceated, Booster.trees_to_dataframe can be used instead to extract tree data from a Booster object.

Explain

decompose_prediction(tabular_trees, row)	Decompose prediction from tree based model with Saabas method[1].
PredictionDecomposition(summary, nodes)	Prediction decomposition results.
calculate_shapley_values(tabular_trees, row)	Calculate shapley values from TabularTrees model for row of data.
ShapleyValues(summary, permutations)	Shapley values results.

Warning

The calculate_shapley_values function is very slow and is only implemeneted for illustration purposes.

Both xgboost and lightgbm implement the must faster treeSHAP algorithm, accessible via the Booster.predict methods when specifying pred_contribs or pred_contrib respectively.

Validate

validate_monotonic_constraints(...[, ...])	Validate that trees conform to monotonic constraints.
MonotonicConstraintResults(summary, ...)	Results of checking monotonic constraints.