摘要
目的利用GM(1,1)、ARIMA和BP神经网络模型探索肾综合征出血热(hemorrhagic fever with renal syndrome,HFRS)预测预警的最优模型,为广州市的HFRS防控预测预警提供科学建议。方法系统收集广州市2006-2019年HFRS发病数据,采用GM(1,1)、ARIMA和BP神经网络模型以及混合模型对2006-2016年发病率进行拟合建立HFRS预测已经模型,并对2017-2019年发病率进行预测,通过平均相对误差选取最佳预测预警模型。结果2006-2019年广州市共报告HFRS病例2453例,年均发病率1.373/10万,2013年发病率最高,为1.794/10万,2019年发病率最低,为0.954/10万。1-4月共报告1106例病例,占病例总数的45.08%。ARMIA(0,1,1)(0,1,1)12模型预测的平均相对误差最小为27.75%,GM(1,1)、BP神经网络、BP-GM和BP-ARIMA模型预测的平均相对误差分别为33.280%、27.750%、29.670%、46.210%和46.600%。结论基于月度数据进行建模的ARIMA模型识别异常值的能力更好,且预测的平均相对误差率最小。
Objective To use utilize GM(1,1),ARIMA and BP neural network models to explore the optimal model of hemorrhagic fever with renal syndrome(HFRS)forecast and early warning,and to provide scientific advices for the forecast and early warning of HFRS in Guangzhou.Methods Systematic collection of HFRS incidence data in Guangzhou from 2006 to 2019 was systematically collected,GM(1,1),ARIMA and BP neural network models and mixed models were used to fit the incidence of 2006 to 2016 and establish a HFRS prediction model,and the incidence rate from 2017 to 2019 was predicted,and the best prediction and early warning models were selected through average relative error.Results A total of 2453 cases of HFRS with an average annual incidence rate of 1.373/100000 were reported in Guangzhou from 2006 to 2019,and the incidence rate hit maximum value in 2013,being 1.794/100000,but the incidence rate was minimum in 2019,hitting 0.954/100000.A total of 1106 cases were reported from January to April,accounting for 45.08%of the total number of cases.The average relative error predicted by the ARMIA(0,1,1)(0,1,1)12 model was minimum,hitting 27.75%,and the average relative error predicted by the GM(1,1),BP neural network,BP-GM and BP-ARIMA models were 33.280%,27.750%,29.670%,46.210%and 46.600%.Conclusion ARIMA model based on monthly data delivers better ability to identify outliers,and average relative error rate of prediction is the minimum.
作者
祁娟
陈守义
陈海燕
徐建敏
许聪辉
王大虎
魏跃红
QI Juan;CHEN Shou-yi;CHEN Hai-yan;XU Jian-min;XU Cong-hui;WANG Da-hu;WEIYue-hong(Guangzhou Center for Disease Control and Prevention,Guangdong510440,China)
出处
《医学动物防制》
2021年第5期485-489,共5页
Journal of Medical Pest Control
基金
广州市科技计划项目(201607010130)。
关键词
肾综合征出血热
预测预警
ARIMA
灰色预测模型
神经网络
Hemorrhagic fever with renal syndrome
Prediction and warning
ARIMA
Grey prediction model
Neural network