The Improved Lstm Method Was Applied To The Predictive Analysis Of Covid-19 Propagation In The GM(1,1) Model

Infectious disease is a kind of disease that spreads between people and animals,between people and animals through different pathogens.The sudden outbreak of infectious diseases will not only cause panic to the society,but also bring irreversible impact on the national economy.Research on the spread of the epidemic can not only predict the future spread trend in advance,but also provide valuable advice for the deployment of prevention and control policies.Aiming at the problem of predicting the number of new daily COVID-19 transmission in China,this paper studies a new infectious disease combination prediction model based on feature selection and deep learning model.First of all,excel was used to collect and sort epidemic data from March 1,2020 to October 31,2022,taking China’s full liberalization as the time node,and pre-processing such as data cleaning,standardization and normalization was done.Feature vector selection methods and prediction models include:variance selection method,Adaptive-lasso regression,random forest,GM(1,1)and LSTM model.A group of most explanatory influencing factors("new deaths","suspected confirmed cases","newly cured cases","total cured cases","total deaths","total confirmed cases","currently suspected confirmed cases" and "total close contacts")were selected.By comparison,it was concluded that GM(1,1)had the best ability to screen variables.The GM(1,1)-LSTM combination prediction model is constructed.Secondly,the GM(1,1)-LSTM model is optimized by introducing SVR model and random search algorithm.A GM(1,1)-LSTM-SVR model was established to make up for the shortcomings of the LSTM model by the advantages of small sample learning and nonlinear mapping of the SVR model.The superparameter combination of the LSTM model was optimized by random search.Finally,the combination prediction model before and after optimization was evaluated by the model evaluation index.It is determined that the optimized combination prediction model has the highest accuracy.In conclusion,this dissertation uses feature vector selection and deep learning to study the prediction and analysis methods of COVID-19 transmission,adding new ideas to the research in this field.A new combined model of epidemic trend prediction,GM(1,1)-LSTM model,was established and optimized to improve the prediction accuracy of the combined model.The combined prediction model established in this paper can not only be used to predict the spread trend of the novel coronavirus,but also be extended to the prediction research of other infectious diseases,and provide reference for the deployment of prevention and control measures in the early stage of the outbreak of infectious diseases.