COVID-19 virus mutation prediction with LSTM and attention mechanisms

Abstract

Abstract Coronavirus disease 2019 (COVID-19), caused by Severe Acute Respiratory Syndrome Coronavirus 2, is an emerging and rapidly spreading type of coronavirus. One of the most important reasons for the rapid spread of the COVID-19 virus are the frequent mutations of the COVID-19 virus. One of the most important methods to overcome mutations of the COVID-19 virus is to predict these mutations before they occur. In this study, we propose a robust HyperMixer and long short-term memory based model with attention mechanisms, HyperAttCov, for COVID-19 virus mutation prediction. The proposed HyperAttCov model outperforms several state-of-the-art methods. Experimental results have showed that the proposed HyperAttCov model reached accuracy 70.0%, precision 92.0%, MCC 46.5% on the COVID-19 testing dataset. Similarly, the proposed HyperAttCov model reached accuracy 70.2%, precision 90.4%, MCC 46.2% on the COVID-19 testing dataset with an average of 10 random trail. Besides, When the proposed HyperAttCov model with 10 random trail has been compared with compared to the study in the literature, the average of performance values has been increased by accuracy 7.18%, precision 37.39%, MCC 49.51% on the testing dataset. As a result, the proposed HyperAttCov can successfully predict mutations occurring on the COVID-19 dataset in the 2022 year.