Comparison of machine learning models for coronavirus prediction

Coronavirus, also known as COVID-19, was first detected in Wuhan, China, December 2019. It is a family of viruses ranging from the common cold to severe acute respiratory syndrome (SARS). The symptoms such virus are similar those or seasonal allergies. Like other viruses, it mainly transmitted through airborne droplets when coughing sneezing. Therefore, recognition COVID-19 requires careful laboratory analysis, and reduction resources major challenge. On 11 March, 2020, World Health Organization (WHO) declared caused by SARS-CoV-2, pandemic, there had been an exponential increase cases worldwide, demand for intensive beds related structures far exceeded existing capacity. examples this regions Italy. Brazil registered case SARS-CoV-2 on 02/26/2020. Transmission country shifted very quickly imported local and, finally, community missions, with Brazilian federal government announcing national transmission 03/20/2020. As March 23, state São Paulo population about 12 million people, where Israelita Albert Einstein Hospital located, 477 disease 30 deaths were registered, 27, already 1223 68 concomitant deaths. To slow spread Paulo, quarantines social distancing measures introduced. One motivations challenge fact that, context extensive healthcare system possible limitation testing, not practical test every case, results can only be used testing target subpopulation. study objective build model based machine learning that predict detection medical data. For this, various classification models compared, best one coronaviruses determined. comparison individuals class 1, i.e., positive test. required determine response F1 score 1. Materials Methods . An open-source data set Brazil, taken basis. following study: RandomForests (RF), K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Logistic Regression (LR), Decision Tree (DT) AdaBoost (AB), well 10-time cross-validation technique. Some performance measures, accuracy, recall, evaluated. Results Out total 5,644 people tested during 5,086 negative 558 positive. At same time, support vectors showed detecting coronavirus recall 75 % 60 compared models: Random drill, KNN, LR, AB, DT. Discussion Conclusions found using AB algorithms, greater accuracy achieved, but stability LSVM algorithm higher. recommended useful tool COVID-19.