Standard methods of using categorical variables as predictors either endow them with an ordinal structure or assume they have no structure at all. However, categorical variables often possess structure that is more complicated than a linear ordering can capture. We develop a mathematical framework for representing the structure of categorical variables and show how to generalize decision trees to make use of this structure. This approach is applicable to methods such as Gradient Boosted Trees which use a decision tree as the underlying learner. We show results on weather data to demonstrate the improvement yielded by this approach.

Exploiting Categorical Structure Using Tree-Based Methods