Tree ensembles

Back to Models and Features

Tree ensembles average or boost many decision trees. They capture nonlinearity and interactions automatically and are the workhorses behind the largest reported macro forecasting gains.

Pass any model string below as Arm(model=...). Extra names an optional dependency, Scaling flags whether predictors should be standardized, and Tunable counts the hyperparameters the search space exposes.

Model string	Description	Input	Extra	Scaling	Recommended preprocessing	Tunable
catboost	CatBoost regressor.	supervised	catboost	no	default	3
decision_tree	CART regression tree.	supervised	none	no	default	2
extra_trees	Extremely randomized trees.	supervised	none	no	default	3
gradient_boosting	Gradient-boosted regression trees.	supervised	none	no	default	3
lgb_plus	LGB+ competition hybrid boosting with tree/linear channel diagnostics.	supervised	lightgbm	no	default	7
lgba_plus	LGB^A+ alternating tree-block and greedy linear boosting.	supervised	lightgbm	no	default	7
lightgbm	LightGBM regressor.	supervised	lightgbm	no	default	4
macro_random_forest	Adapter for the external MacroRandomForest package.	supervised	macro_random_forest	no	default	8
quantile_regression_forest	Quantile regression forest.	supervised	none	no	default	3
random_forest	Random forest regression.	supervised	none	no	default	3
xgboost	XGBoost regressor.	supervised	xgboost	no	default	4

Reference

• Models reference page for ModelSpec, ModelFit, and fit conventions.