Abstract
This paper presents a study on the diagnosis of COVID-19 through the analysis of biomarkers from clinical tests using Machine Learning techniques. The research was conducted in two stages. In the first, the influence of clinical markers on the diagnosis of the disease was investigated, and in the second, the performance of several Machine Learning algorithms was investigated in the classification of patients with symptoms similar to COVID-19.
The experiments used a database provided by Hospital Israelita Albert Einstein with laboratory test results from 5,644 patients submitted to the RT-PCR test. Among the results found in the variable selection stage, indicators from the leukocyte group were more relevant for COVID-19 detection. In the classification step, the best results were obtained for the
The results were obtained with Stacking using 20 descriptors selected by Decision Tree (Acuracy = 0.9778; Sensitivity = 0.9527; Specificity = 1.000). The results indicate
it is feasible to use Machine Learning techniques together with variable selection to obtain models with good predictive power for the diagnosis of COVID-19.
DOI: https://doi.org/10.56238/colleinternhealthscienv1-023