MACHINE LEARNING TECHNIQUES FOR PREDICTION OF COVID-19 IN POTENTIAL PATIENTS

Abstract

The coronavirus pandemic overwhelmed many countries and a shortage of testing kits and centers for exposed patients worsens the situation in most countries. These have prompted the need to quickly predict COVID-19 in patients and stop the spread of the virus. In this research, we present a method for predicting COVID-19 based on symptoms, and to make this system efficient, the dataset was obtained from Afriglobal Laboratory Nigeria, and preprocessing and feature extraction were done on the dataset. Three classifiers, logistic regression, support vector machines, and hybridization of the logistic regression and support vector machines were used to train the data. The test data were evaluated against the model, and the research found that the performance analysis values for accuracy, precision, recall, and F1score for logistic regression (LR) are 91%, 91%, 95%, and 93%, for Support Vector Machines (SVM), 94%, 93%, 100%, and 96% and for the Hybridized model (LR+SVM) are 95%, 94%, 98%, and 96%.  To get the parameters needed for the performance evaluation of the classifiers, the confusion matrix method was employed. In comparison to existing methods and studies, the hybridized system performs better than LR and SVM models. As a result, the hybridized model can accurately predict Covid-19.