Análise de sinais e sintomas da infecção pelo vírus SARS-CoV-2 considerando diferentes ondas utilizando aprendizado de máquina

Abstract

In March of 2020, the World Health Organization, declared the world pandemics of COVID-19, witch can manifest in humans in consequences of virus infection of SARS­ CoV-2. This work consists in use techniques of Data Mining and Machine Learning for the infections diagnosis. A methodology was created to facilitate this task and can be ap­ plied in any outbreak or pandemic waves. The method, in addition to generate diagnoses models based only signals and symptoms, can evaluate if there is differences of signals and symptoms between waves (or outbreaks), through of Explainable techniques of the models of Machine Learning. Another aspect, It is identify possible differences of quality between exams, for example, Rapid Test and RT-PCR. The study of case in this work, based on patient data who sought care on Piquet Carneiro Polyclinic of State University of Rio de Janeiro. In this work, the results obtained with the tests was used as diagnoses of symptomatic infection on SARS-CoV-2 virus, in addition of related signals and symp­ toms, and date of initial of this signals and symptoms. Using the Random Forrest model, it was possible achieve the result of up to 76% of Sensitivity, 86% of Specificity and 79% of accuracy in the results oftests in one contagion wave of SARS-CoV-2 virus. It was pos­ sible identify differences of signals and symptoms between contagion waves, in addiction to the observation that exams of RT-PCR is more accurate than Rapid Test of Antibodies.