Research On Virus Propagation Process Based On Deep Neural Network

There have been many outbreaks of infectious diseases in human history,such as cholera,black death,smallpox virus,etc.they are highly infectious,and each outbreak is a catastrophe in human history.Nowadays,with the deterioration of climate and environment,the acceleration of human migration is more conducive to the spread of the virus.Therefore,it is critical to prevent and control the spread of the virus timely.Among them,the most well-known infectious disease prediction model is the warehouse model.Now most scholars make improvements based on the classic warehouse model.in order to achieve better prediction effect.However,in the process of implementation,it is found that the more complex the model is,the more refined the data is needed.For example,it is very difficult to determine the number of asymptomatic infected persons and the number of days of incubation period.This leads to inaccurate prediction results.In order to better study the spread process of virus,this paper establishes the traditional time series data prediction model ARIMA model,deep neural network LSTM model and cnn-lstm model with convolution layer for analysis and comparison.Combined with the outbreak of new coronavirus SARS at the end of 2019 The novel coronavirus pneumonia epidemic data caused by Co V-2 are based on the COVID-19 data from four countries in China,Italy,the United Kingdom and Germany.The improved SIS model with nonlinear infection rate,the prediction model based on the differential integration moving average autoregression（ARIMA）,the long and short term memory artificial neural network（LSTM）and the LSTM network set with the volume layer are established respectively.LSTM prediction model is used to simulate and predict the development of epidemic situation in four countries.By calculating the prediction error to evaluate the model,it is concluded that LSTM model can well predict the spread of virus.