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。
Abstract
Objective To explore the application of BP neural network and SARIMA model in the prediction of the incidence 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 January 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 2 923 cases of HFMD were reported in Donghu District,with an average annual incidence rate of 52.56/100 000.The incidence rate of HFMD was different in different years (χ2=286.066,P<0.001).For the prediction 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
Key words
BPNN model /
SARIMA model /
Hand-foot-mouth disease /
Prediction
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 李颉,郑步云,王劲峰.2008—2018年中国HFMD时空分异特征[J].地球信息科学学报,2021,163(3):419-430.
[2] 于勇.北京市手足口病与气象因素关联性研究及时间序列模型预测[D].沈阳:中国医科大学,2020.
[3] 刘文东,吴莹,艾静,等.BP神经网络在痢疾发病趋势预测中的应用研究[J].中国卫生统计,2012,29(6):801-804.
[4] 刘天,姚梦雷,黄继贵,等.BP神经网络在传染病时间序列预测中的应用及其MATLAB实现[J].预防医学情报杂志,2019,35(8):812-816.
[5] 陈春艳,陈亿雄,赵执扬,等.SARIMA模型和LSTM神经网络在预测深圳市宝安区手足口病疫情中的应用[J].山西医科大学学报,2022,53(10):1302-1307.
[6] 周浩,李虹,张岳琴,等.SARIMA模型在山西省手足口病发病预测中的应用[J].中国预防医学杂志,2023,24(2):117-121.
[7] 张京凤.基于ARIMA和BPNN模型的全国手足口病发病率预测[D].重庆:西南大学,2021.
[8] 曹飞,李辉,刘明斌,等.2008—2012 年南昌市手足口病的流行特征及时空聚集性分析[J].现代预防医学,2015,42(8):1356-1373.
[9] 王晓文,施勇,李健雄,等.江西省2009—2017年手足口病流行病学特征及病原学研究[J].病毒学报,2018,34(6):844-849.
[10] Tian CW,Wang H,Luo XM.Time-series modelling and forecasting of hand,foot and mouth disease cases in China from2008 to 2018[J].Epidemiol Infect,2018,147(6):82-84.
[11] 陶君雯,张韬,庄雪菲,等.动态贝叶斯网络模型和SARIMA模型对手足口病预测效果的比较[J].现代预防医学,2020,47(21):3851-3854.
[12] 郝启迪,米莎莎,李凡卡.利用ARIMA和BP神经网络模型分析新疆生产建设兵团乙肝流行特征[J].现代预防医学,2021,48(18):3265-3269.
[13] 杨文姣,肖俊玲,丁国武.ARIMA模型和BP神经网络模型在甘肃省结核病发病率预测中的应用[J].中华疾病控制杂志,2019,23(6):728-732.
[14] 刘辉,马殿梅,刘晓坤,等.应用ARIMA-BPNN组合模型预测手足口病发病率[J].现代预防医学,2016,43(16):2885-2888.
[15] 吴文博,李虹艾,万鹏程,等.利用遗传算法优化的ARIMA-BP组合模型预测手足口病发病趋势[J].中南医学科学杂志,2014,42(6):572-576.
基金
江西省卫生健康委科技计划(202311421)
PDF(1140 KB)
