基于SEIR-LSTM混合模型的新型冠状病毒肺炎传播趋势分析与预测  

Analysis and prediction of the spread trend of COVID-19 based on SEIR-LSTM mixed model

作　　者：洪彬 陈锦秀[1] 王连生[1] 俞容山 HONG Bin;CHEN Jinxiu;WANG Liansheng;YU Rongshan(School of Informatics,Xiamen University,Xiamen 361005,China)

机构地区：[1]厦门大学信息学院,福建厦门361005

出　　处：《厦门大学学报（自然科学版）》2020年第6期1034-1040,共7页Journal of Xiamen University：Natural Science

基　　金：国家自然科学基金(61671399);中央高校基本科研业务费专项(20720190012)。

摘　　要：为了对新型冠状病毒肺炎(COVID-19)传播趋势实现更精准的分析与预测,基于传统的传染病动力学模型SEIR(susceptible-exposed-infectious-recovered)和长短期记忆(long short-term memory,LSTM)深度学习网络构建了一种混合模型SEIR-LSTM,结合全国和湖北省的疫情数据进行分析与预测.从实验结果来看,相较于传统的机器学习、多项式拟合、普通的SEIR模型和单一的LSTM,所提出的混合模型能取得更高的疫情预测精度.最后给出了对疫情发展的预测,对疫情防控具有一定的应用价值和社会价值.To achieve more accurate analyses and predictions of the spread trend of the Coronarirus Disease 2019(COVID-19),we have developed a hybrid model SEIR-LSTM based on the traditional dynamic model SEIR and deep learning network LSTM,as well as analyze and predict epidemic data of the country and Hubei Province.From these results,compared with traditional machine learning,polynomial fitting,ordinary SEIR model and single LSTM,the mixed model can achieve higher accuracy.Finally,the prediction of the development of the epidemic situation is given,possibly contributing to the prevention and control of the epidemic situation.

关 键 词：新型冠状病毒肺炎 SEIR模型 长短期记忆 混合模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]