Aplicação de machinne learning para predição da evolução dos casos de COVID-19 no Brasil após a vacinação

Abstract

The COVID-19 pandemic, caused by the SARS-CoV-2 virus, led to an unprecedented global crisis, affecting millions of people and putting immense pressure on healthcare systems worldwide. In Brazil, the impact was particularly severe, with high rates of infection and mortality. In this context, the implementation of effective strategies to predict the progression of COVID-19 cases became crucial. This study aims to analyze the efficacy of different machine learning models in forecasting the progression of COVID-19 cases in Brazil, with a focus on mortality rates following the onset of the vaccination campaign. The analysis utilized epidemiological data from official sources, considering patients' pre-existing conditions and vaccination status, while excluding symptomatic and clinical progression variables. Data mining and machine learning techniques were applied to classify the variables, using Decision Trees, Random Forests, XGBoost, Multi-Layer Perceptron, and KNN models. Model performance was evaluated through accuracy, precision, recall, and F1-Score metrics. Among the results, the Random Forest model performed the best, achieving an average F1-Score of 63% in classifying both classes. The importance of the variables was assessed using the SHAP library, identifying age and vaccination as the most significant factors for model classification. The study demonstrated that machine learning can provide valuable insights for combating infectious diseases and highlighted the potential need for including symptomatic and clinical variables to improve the performance in classifying the target variable.