Research On Estimation Methods Of Population Activity Patterns For Large-scale Infectious Diseases

The estimation of human activity patterns in the real world can significantly improve the predictive ability of infectious disease models in complex social environments,especially for the prediction of the spread of large-scale infectious diseases worldwide.In this paper,oriented to modeling and predicting large-scale infectious diseases,based on network science and statistical methods,combined with the big data information available to the society,we studied two important population activity patterns and proposed corresponding estimation methods.The specific research work includes the following three aspects:First,we analyzed the theoretical basis of using Bayesian hierarchical model to infer the human contact pattern from the survey data of individual’s contact,and gave the modeling method for estimating the posterior distribution of model parameters based on Bayesian generalized linear regression.Based on the method,we used the survey data conducted by Fudan University to estimate the contact pattern in Shanghai,and obtained the contact matrix of home,work,school and other location.Secondly,we proposed an algorithm for estimating the human mobility network based on bipartite graph.The algorithm integrates location data and flight data,and estimates the human travel flow between all prefectures through network science methods such as breadth-first traversal.We used this method to estimate the travel flows between prefectures in mainland China from January to April 2020,and built a prefecture-level human mobility network.Thirdly,we simulated the spread of epidemic infectious diseases using the estimated human activity pattern.Based on the age-structured SIR model incorporating the human contact pattern,we conducted simulation experiments on the spread of infectious diseases,and observed the positive effect of the mean contact rate on the infection rate of diseases.In addition,we simulated the early spread of COVID-19 in mainland China based on the metapopulation model incorporating the human mobility pattern,and the results showed that the use of human mobility network could better model and predict the spatial spread of infectious diseases.