Why can two individually weak categorical columns still matter to CatBoost even without manually adding their…
Classic MLLevel: IntermediateCatBoost · Categorical statistics
Why can two individually weak categorical columns still matter to CatBoost even without manually adding their concatenation?
Hints
- Joint and marginal category information differ.
- Combinations increase both expressiveness and sparsity.
Answer
CatBoost can construct statistics on feature combinations. A combination can capture information absent from marginal categories, but higher cardinality reduces support and requires validation rather than assuming every interaction helps.
Source and licence: CatBoost · Categorical statistics · Apache-2.0 · answer: Quantoros editorial