Biological age is meant to reflect how quickly a person is aging relative to their chronological age. Machine-learning models have improved biological age estimation, but many rely mainly on continuous laboratory biomarkers, potentially overlooking categorical information such as medical history or lifestyle factors.
The new arXiv preprint proposes using CatBoost, a gradient-boosting algorithm, together with SHAP for interpretability, trained on data from NHANES. By including both continuous and categorical features, the authors aim to make biological age estimates more clinically interpretable, rather than treating the model as a black box.
As a preprint, the work has not yet been peer-reviewed, and the abstract does not report performance metrics or detailed methodology. Still, the explicit focus on categorical features and explainability addresses a recognized gap in the field, though the practical benefits will require closer scrutiny.