BP神经网络和SARIMA模型在手足口病发病率预测中的比较  

Comparison of BP neural network and SARIMA model in prediction of hand-foot-mouth disease

作　　者：李剑 刘莉红[2] 何志勇 徐幼莉 叶珍珍 LIJian;LIULihong;HEZhiyong;XUYouli;YEZhenzhen(Donghu District Comprehensive Health Supervision and Law Enforcement Bureau,Nanchang 330006,Jiangxi,China;Yichun City Center for Disease Control and Prevention,Jiangxi 336000,China;Donghu District Center for Disease Control and Prevention,Nanchang 330006,Jiangxi,China;Jiangxi Provincial Center for Disease Control and Prevention,Nanchang 330029,Jiangxi,China)

机构地区：[1]东湖区卫生健康综合监督执法局,江西南昌330006 [2]宜春市疾病预防控制中心,江西336000 [3]东湖区疾病预防控制中心,江西南昌330006 [4]江西省疾病预防控制中心,江西南昌330029

出　　处：《安徽预防医学杂志》2023年第6期454-458,496,共6页Anhui Journal of Preventive Medicine

基　　金：江西省卫生健康委科技计划(202311421)。

摘　　要：目的 探讨BP神经网络和季节差分自回归移动平均模型(SARIMA)模型在手足口病(HFMD)发病率预测中的应用,建立HFMD发病的预警系统,为其有效防控提供参考依据。方法 收集2012-2020年南昌市东湖区HFMD的发病数据,用BP神经网络和SARIMA进行模型拟合,预测2021年1-12月HFMD的发病情况,并用MSE、MAE、MAPE来评估2种模型的预测效果。结果 2012-2021年东湖区累计报告HFMD例2 923例,年平均发病率为52.56/10万,不同年度HFMD发病率有所不同(χ2=286.066,P<0.001)。对于2021年1-12月东湖区HFMD发病率的预测,BP神经网络模型的MSE、MAE、MAPE分别为2.87、1.37、0.82%,SARIMA模型MSE、MAE、MAPE分别为11.47、2.85、6.24%。结论 BP神经网络和SARIMA模型均适用于预测HFMD发病率,且BP神经网络模型的预测效果优于SARIMA模型。Objective To explore the application of BP neural network and SARIMA model in the prediction of the inci-dence rate of hand-foot-mouth disease(HFMD),to establish an early warning system for HFMD,and to provide a reference for its effective prevention and control.Methods Data of HFMD incidence in Donghu District,Nanchang were collected from 2012 to 2020.BP neural network and SARIMA were used for model fitting to predict the incidence of HFMD from Jan-uary to December 2021.MSE,MAE,and MAPE were used to assess the predictive effectiveness of the 2 models.Results From 2012 to 2021,a total of 2923 cases of HFMD were reported in Donghu District,with an average annual incidence rate of 52.56/100000.The incidence rate of HFMD was different in different years(χ^(2)=286.066,P<0.001).For the predic-tion of incidence rate of HFMD in Donghu District from January to December in 2021,the MSE,MAE,and MAPE of BP neural network model were 2.87,1.37,and 0.82%,respectively.While the MSE,MAE,and MAPE of SARIMA model were 11.47,2.85,and 6.24%,respectively.Conclusion Both BP neural network and SARIMA model can be used to predict the incidence rate of HFMD.The prediction effect of BP neural network model is better than SARIMA model.

关 键 词：BP神经网络模型 SARIMA模型 手足口病 预测 

分 类 号：R181.8[医药卫生—流行病学]