摘要
从机理建模和数据建模出发,结合猩红热2012~2017年发病率数据,分别建立了基于SIS模型近似解析解的改进模型,基于NAR神经网络的猩红热传染病预测模型,以及基于局部加权线性回归的猩红热传染病预测模型,用以描述猩红热的传播过程。运用所建立的三种模型分别对2018年1-6月份猩红热发病率进行预测,计算各模型误差以及和方差,得出相对于近似解析解模型,改进模型以及神经网络模型而言,局部加权线性回归模型能更加精确地描述猩红热传染病的发病趋势。
Starting from the perspective of mechanism modeling and data modeling,and combined with the incidence data of scarlet fever from 2012 to 2017,this paper respectively established an improved model based on approximate analytical solutions of SIS model,a scarlet fever epidemic prediction model based on NAR neural network,and a scarlet fever epidemic prediction model based on local weighted linear regression.The established three models were used to respectively predict scarlet fever incidence from January to June,2018.By calculating errors and variances of each model,it is concluded that the local weighted linear regression model can describe the epidemic trend of scarlet fever more accurately than the approximate analytical solution model,the improved model and the neural network model.
作者
刘晓东
魏海平
曹宇
魏宇峰
LIU Xiao-dong;WEI Hai-ping;CAO Yu;WEI Yu-feng(Liaoning Petrochemical University,Fushun Liaoning 113001,China;China Petroleum Corporation Hebei Zhangjiakou Sales Company,Zhangjiakou 075000)
出处
《计算机仿真》
北大核心
2020年第8期171-176,375,共7页
Computer Simulation
基金
辽宁省教育科学‘十三五’规划立项课题(JG18DA031)。
关键词
模型
数据拟合
神经网络
局部加权线性回归
Model
Data fitting
Neural network
Local weighted linear regression
