Predicción de uso de servicios de salud en empresas aseguradoras, mediante KDD y Aprendizaje Automático

Abstract

Abstract: In this work, a predictive model was developed with the aim of anticipating the use of medical services in insurance companies, using machine learning techniques. The data analysis was structured following the KDD methodology, which organizes the process in clear stages. The "Health Insurance" dataset served as a basis for training the models, and they were evaluated using key metrics, which guaranteed the accuracy and reliability of the predictions. A predictive model was developed along with the creation of an interactive dashboard using the Plotly Dash library, which allows a clear and detailed visualization of the results obtained, highlighting that this dashboard offers users the possibility of comparing the predictions of the model with the real values ​​and segmenting the results according to key variables, such as the sex and tobacco consumption of the insured. The tool not only facilitates the interpretation of the results, but also allows interactive exploration of different factors that influence the use of medical services. Together, the predictive model and the dashboard provide a solution for insurance companies to improve the management of their resources. The implementation of these tools allows them to anticipate the demand for medical services with greater precision, which optimizes the planning and allocation of resources, and improves operational efficiency within the insurance sector.