The Study On The Development And Application Of High Dimensional Self-learning Simulator For COVID-19

The rapid spread of SARS-Co V-2 has caused a huge impact on national health and economic development in China.Accurate prediction of the risk and the future development trend of COVID-19 in provinces and municipalities is particularly important for targeted prevention and control and rapid containment of the epidemic spread.The basic mathematical models of infectious diseases,such as the susceptible-exposedinfectious-recovered(SEIR)model,can predict the trend of the epidemic situation qualitatively under ideal conditions.During the outbreak,given existence of phenomena and actual situations such as domestic large-scale population migrations across the regions,a variety of prevention and control measures,personal protective consciousness change,medical and health system limit bearing capacity,and super-spreading events,the above factors may be far-reaching influence on epidemic situation.The traditional infectious disease model is difficult to make accurate predictions of epidemic.Therefore,it is hard to provide support for precise prevention and control decisions.In this study,a high-dimensional self-learning COVID-19 epidemic simulator was built by using machine learning technology based on the confirmed case information,population migration,medical resources and other data.A variety of transmission and prevention and control factors were taken into consideration.The simulator supports multi-level granularity of province,city,district and county.It can provide accurate reconstruction,scenario simulation and prediction of the epidemic trend,thus assisting in situation analysis and evaluating the effectiveness of prevention and control decisions.The simulator uses a complex transmission dynamics model(SHEADIOR model)to represent the changes of population health status.Through the input of real-time and accurate population migration data,the inter-regional and intra-regional population movement and social contact patterns are depicted.The optimal distribution of virus transmission characteristic parameters was fitted by using detailed case information of COVID-19.Using the data of first stage epidemic in China as the test set,the full pattern of transmission dynamic was reconstructed and verified by simulator at the provincial level.This provides a simulation experimental environment for the subsequent evaluation of prevention and control policies and the simulation of hypothetical scenarios.By changing the time point and parameters of the prevention and control measures,the effects of key measures such as city closure,delay of returning to work and school reopening and national medical support to Hubei were evaluated.The simulation results show that city closure in Wuhan is very timely.It can significantly reduce the estimated number of infected individuals in Hubei Province and the whole country.The risk of a premature and full resumption of work and schools reopening is very high.Unless strict intervention measures are implemented again,it will be difficult to control the epidemic.National support to Hubei is necessary,otherwise the epidemic in Hubei province will not be brought under control before June.Governments around the world have dealt with the COVID-19 in different ways.Through simulations,comparative assessments were made on the models of prevention and control in different countries.According to the simulation results,if China had adopted the prevention and control model of other countries,it is estimated that by April30,there would have been more than 100,000 confirmed cases.British "herd immunity" policy and American loose measures are expected to lead to more than a million infections.On the contrary,if the UK and the US adopt strict prevention and control measures similar to those taken in Hubei Province,the COVID-19 outbreak is expected to be brought under control within three months.The number of confirmed cases is expected to decrease significantly.The outbreak could be over by mid-May if Italy can seal off the worst-hit areas in the north 20 days earlier.At the same time,the simulation results show that the application effect of the "Korean model" without city closure but multi-detection is not ideal in China.Without controlling population activity and keeping social distancing,solely relying on nucleic acid testing and improving personal protection in public places,the expected number of infections will be higher,suggesting the need for closed management of high-risk areas in the early stages of the outbreak.Since June,the COVID-19 situation in China has been in a low epidemic state,but it is facing the risk of small outbreak of point-origin.Combined with the epidemiological survey data and the modified simulator model,this study reviewed and analyzed the transmission dynamics of COVID-19 in Beijing Xinfadi Wholesale Market,and simulated and evaluated a series of targeted interventions.The simulation results show that targeted interventions,including localized lockdown intervention,close contact tracing,and community-based nucleic acid testing,are sufficient to effectively control the outbreak,thus achieving the optimal balance between epidemic control and economic development.Western countries are experiencing surges in COVID-19 cases and deaths due to increasing public transportation during holiday seasons.This study explored whether the COVID-19 would rebound in China during the Spring Festival travel rush,and optimized the nucleic acid testing strategy to contain the epidemic.The local transmission of the epidemic during the Spring Festival travel rush under different nucleic acid testing strategies was predicted by using a microsimulation model and COVID-19 simulator.The total number of infections,the burden of nucleic acid testing and the overall medical expenditure were calculated to optimize the nucleic acid testing strategy during the Spring Festival holiday.The simulation results show that,compared with no nucleic acid testing,testing passengers from medium-high risk regions 3 days before travelling could significantly reduce the risk of transmission,and it is more economical and efficient than testing for all passengers.