| At present,more in-depth research results have been achieved on the modeling and analysis technologies for the spread prediction and control of the COVID-19 epidemic,such as the compartment model,multi-agent modeling and machine learning algorithms.However,the COVID-19 prediction models generated by these types of modeling technologies have many defects.Firstly,limited by computing resources and data scale,it is difficult to accurately simulate the impact of individual mobility in different regions on the epidemic and prevention and control measures;secondly,the current research lacks consideration of the complex situation of some epidemics,such as the imbalance of vaccine coverage caused by differences in medical resources in various countries in the future,and the emergence of new mutant strains.Therefore,this topic will focus on the normalization of COVID-19 prevention and control caused by the contradiction between strain variation and vaccination effectiveness,how to use multi-regional individual mobility modeling technology to achieve COVID-19 transmission prediction and efficient and accurate prevention and control strategy support,etc.This thesis proposes an infectious disease prediction and prevention and control model based on individual mobility behavior.The main contributions are as follows:(1)By establishing the relationship between the individual mobility intensity and the effective reproduction number of the epidemic,the parameters of the effective reproduction number of the epidemic are predicted according to the change of the mobility intensity of the population,which is used as the input of the next stage of the compartment model,which enhances the compartment model’s ability to simulate realistic uncertainties.capabilities and improve forecast accuracy.And a new MFConv LSTM ensemble model based on the Bagging method is proposed,which uses multi-scale features(time domain features,frequency domain features,time domain local features,etc.)to construct multiple predictors,which can capture the characteristics of human individual routine activities from multiple perspectives,analyze the dynamic changes of individual mobility in different regions under the epidemic situation as comprehensively as possible,and better auxiliary prediction.(2)Construct an infectious disease compartment prediction model that can adapt to different regional population data,medical resources,multiple mixed strain infections,vaccination and other factors,which can more comprehensively simulate the dynamic change process of epidemic transmission and meet the needs of Adapting to the multifaceted needs of public health decision makers,such as medical resource needs,vaccination plans,etc.,provides interpretable epidemiological theoretical support for the overall prediction model.(3)Established an evaluation of the effect of different prevention and control policies,that is,changing the impact of liquidity in different regions on epidemic prevention and control,and proposed the most efficient and accurate prevention and control strategies adapted to the United Kingdom.The optimal intervention strategy is to implement inhibitory intervention with P=20%(intervention intensity: the liquidity is reduced to 20%of Baseline)before the end of March 2021,and then gradually release the intervention intensity P+20% every month,and finally in 2021 Release the intensity to P=80% by the end of June.The total death toll would then be 129,041.Such a strategy would enable a trade-off between new infections,medical resource needs and economic losses.And this strategy is similar to the strategy implemented in April in the UK,confirming the effectiveness of the intervention strategy proposed in this study.Compared with the official statistics released by the UK government on June 30,2021,the model can predict new cases with an R2-SCORE of 0.8801 and a relative error rate of less than 5% for the total number of deaths.In addition,the model is also suitable for data from other areas in the world. |
