Parameter Inversion And Prediction Of COVID-19 Model

The outbreak of the novel coronavirus pneumonia（COVID-19）in December 2019.Un-til January 8,2023,the National Health Commission officially defined pneumonia infected by the novel coronavirus as a legal Class B infectious disease.In the past three years,the epidemic has seriously affected human life,health and social development,and had a huge impact on human body and mind.Therefore,in the context of pandemic,it is particularly important to study and explore the transmission mechanism and epidemic trend of COVID-19 by using relevant theories of infectious disease models.Based on the establishment of COVID-19 infectious diseases,this article conducted the following studies:1.The parameter inversion problem of COVID-19 model was studied.First,in view of the spread of COVID-19,a SEAIRD infectious disease model with asymptomatic infected persons was established based on the existing SEIRD model.Then,the COVID-19 epidemic data in Russia was selected,and the parameters of the SEAIRD model were inverted by the optimized genetic algorithm.The model was simplified to minimize the objective function problem.The genetic algorithm based on the elite retention strategy was used to obtain the parameter solution₆)=（0.0116,0.052,0.5362,0.1623,0.055,0.2035）of the SEAIRD model.Finally,the sensitivity analysis of these parameters was further carried out.2.The prediction of the number of COVID-19 infections and deaths by the mixed predic-tion model was studied.First,the SEAIRD-LSTM hybrid prediction model was proposed by linear regression of the prediction results of the SEAIRD model and the LSTM mod-el.Then,the number of COVID-19 infections and deaths in Russia was predicted,and the hybrid prediction model of SEAIRD-LSTM was compared with polynomial regression,L-ogistic regression,SEIR and LSTM models.The results show that the hybrid prediction model of SEAIRD-LSTM achieves good prediction effect under RMSE,MAE,MAPE and²evaluation indexes.Finally,the same method is used to invert and predict the parameters in Germany and the United Kingdom,and the error rate of the real value and predicted value is compared.The feasibility of SEAIRD model and the accuracy of SEAIRD-LSTM hybrid prediction model are verified.