Tree-Structured Boosting: Connections Between Gradient Boosted Stumps and Full Decision Trees
This work addresses the need for interpretable and high-performing models in high-stake domains like medicine by connecting previously isolated additive and interaction models.
The paper introduces tree-structured boosting, a technique that creates a single decision tree to bridge the gap between gradient boosted stumps and full decision trees (CART), showing that varying a single parameter can produce models equivalent to either extreme or hybrid forms that outperform both in predictive performance.
Additive models, such as produced by gradient boosting, and full interaction models, such as classification and regression trees (CART), are widely used algorithms that have been investigated largely in isolation. We show that these models exist along a spectrum, revealing never-before-known connections between these two approaches. This paper introduces a novel technique called tree-structured boosting for creating a single decision tree, and shows that this method can produce models equivalent to CART or gradient boosted stumps at the extremes by varying a single parameter. Although tree-structured boosting is designed primarily to provide both the model interpretability and predictive performance needed for high-stake applications like medicine, it also can produce decision trees represented by hybrid models between CART and boosted stumps that can outperform either of these approaches.