DATA COLLECTION PROTOCOL FOR COVID-19 SEVERITY PREDICTION USING MACHINE LEARNING

This work investigated the efficiency of using data for Covid-19 prognosis through Machine Learning (ML) models and proposed a collection method that generates a tidy dataset for use in ML. Data from Hospital Sírio-Libanês were used, which required an extensive series of pre-processing for the data to become adequate. After these steps and application in ML, a good performance was observed in the KNN and SVM algorithms, with AUC=0.81. Therefore, knowing that the dataset is tidy and efficient, a data collection format was determined that allows for faster use in ML models, without the need to carry out several pre-processing steps, enhancing the handling in the patient care and health resources.