基于长短期记忆网络的半参数SEIR模型  

Semi-parametric SEIR model based on LSTM neural network

作　　者：张静[1] 金彤 ZHANG Jing;JIN Tong(School of Mathematics and Statistics,Hainan Normal University,Haikou 571158,China)

机构地区：[1]海南师范大学数学与统计学院,海南海口571158

出　　处：《东北师大学报(自然科学版)》2025年第1期46-52,共7页Journal of Northeast Normal University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(12161029,11701127);海南省自然科学基金资助项目(121RC149)。

摘　　要：提出了带有非线性传播函数的半参数SEIR模型以捕获疾病的传播,从理论上分析了模型的基本性质及基本再生数.以新冠感染为例,比较了各国疫情初期的传播函数,得出不同地区人口、防疫措施等因素对疫情传播的影响不同.以印度为例,利用长短期记忆(LSTM)神经网络对传播函数的离散值进行了拟合,代回半参数SEIR模型后预测出感染人数,所得结果与经典SEIR模型比较,平均绝对百分比误差降低71.73%.因此,半参数SEIR模型对疫情的理论估计更符合实际情况.A semi-parametric SEIR model with a nonlinear transmission function is proposed to capture the spread of disease.The basic properties and the reproductive number of the model are theoretically analyzed.Taking the COVID-19 as an example,comparing the transmission functions of various countries in the early stage of the epidemic,it is concluded that the impact of factors such as population and epidemic prevention measures on the spread of the epidemic varies in different regions.Taking India as an example,the long short-term memory(LSTM)neural network is used to fit the discrete values of the transmission function,and the number of infections is predicted after substituting the semi-parametric SEIR model,and the average absolute percentage error is reduced by 71.73%compare with the classical SEIR model.Therefore,the semi-parameter SEIR model provides a theoretical estimation of the epidemic that is more consistent with the actual situation.

关 键 词：SEIR模型 传播函数 半参数 长短期记忆神经网络 新冠感染 

分 类 号：O29[理学—应用数学]