Modeling recurrence of COVID-19 and its variants using recurrent neural network

Abstract

Coronavirus disease 19 (COVID-19), a disease caused by severe acute respiratory syndrome-coronavirus-2 (SARS-CoV-2), began as the flu and gradually developed into a highly infectious global pandemic leading to the death of over 6 million people in about 200 countries of the world. Its pathogenic nature has qualified it as a deadly disease, causing moderate and severe respiratory difficulty in infected individuals with the ability to mutate into different variants of the first version. As a result, different government agencies and health institutions have sought solutions within and outside the clinical space. This paper models COVID-19 possible recurrence as variants and predicts that the subsequent waves will be more severe than the first wave. Long short-term memory network (LSTM) was used to predict the future occurrence of COVID-19 and forecast the virus's pattern. Machine evaluation was performed using precision, recall, F1-score, an area under the curve (AUC), and accuracy evaluation metrics. Datasets obtained were used to test the data. The collected characteristics were passed on to the system classification network, demonstrating the function's value based on the system's accuracy. The results showed that the COVID-19 variants have a higher disastrous effect within three months after the first wave.