Aplicação de técnicas automatizadas de aprendizagem de máquina para classificação de óbitos em pacientes com covid-19.

Abstract

This study investigated the application of automated machine learning techniques in classifying cases with a potential risk of death among COVID-19 patients, using a histo- rical database from Fiocruz. This database contained data from COVID-19 patients in three different periods, corresponding to different phases of the disease’s circulation, classifying patients who succumbed to the disease. The aim was to apply AutoML tools adapted to this classification scenario and to evaluate their use and potential for research in the field of clinical analysis. By implementing well-established AutoML tools in the literature, such as H2O AutoML, AutoGluon and PyCaret, this study demonstrated how these technologies can simplify complex analytical processes, highlighting the relevance of these tools in making advanced machine learning techniques accessible to non-specialists and their efficiency in processing clinical data. Although challenges related to the processing time of some to- ols were identified, it was concluded that AutoML has significant potential to facilitate and enhance the application of machine learning models in clinical analytics.