Modelo de ensamble para predecir el diagnóstico de pacientes con sospecha de Covid-19

Abstract

The Covid-19 pandemic has affected millions of people around the world, being described by the World Health Organization as a crisis of global concern. This has generated the need to perform a timely prediction of the diagnosis of patients with high risk of clinical deterioration in medical establishments. The aim of this study is to design and compare the performance of machine-assembly-based machine learning models to predict patients with suspected Covid-19. The research follows the positivist paradigm, quantitative approach, observational design without intervention, predictive level. To carry out this study, 1,000 patient records from two health facilities in Peru were collected. The construction of the models was based on assembly algorithms, such as Random Forest, Extra Trees, Gradient Boosting and AdaBoosting. When comparing the models in terms of accuracy, which measures the percentage of cases correctly classified as patients with suspected Covid-19, a 97% accuracy was obtained for the models based on Random Forest and Gradient Boosting. In addition, Cohen's Kappa value was 0.95, which indicates a very good agreement between the model prediction result and the actual data.