Estimación de gastos médicos para clientes corporativos en beneficio a una agencia de seguros mediante el uso de modelos de aprendizaje automático

Abstract

CONDITION FOR PUBLICATION OF PROJECT. project developed a predictive model to estimate medical expenses covered by insurance companies, using historical data from employees of corporate clients. Through preprocessing techniques, dimensionality reduction (UMAP), clustering (KMeans), and advanced regression models such as Random Forest, Gradient Boosting, and XGBoost, a scalable and accurate tool was built. The model enables more precise forecasting of future medical expenses, improving financial planning for both insurers and policyholders by reducing information asymmetry in risk management. Additionally, an interactive visualization tool was developed to help end users interpret the results with ease. The methodology is proven to be replicable and shows clear opportunities for improvement, including the incorporation of new explanatory variables and the use of more sophisticated algorithms. Future versions of the model should integrate larger datasets to enhance the prediction of specific health conditions.