2017—2022年浙江省其他感染性腹泻病ARIMA模型预测精度分析  

Analysis of the prediction accuracy of ARIMA model for other infectious diarrhea diseases in Zhejiang Province from 2017 to 2022

作　　者：茅蓉[1] 向泽林[1] 王远航 陆许贞 MAO Rong;XIANG Zelin;WANG Yuanhang;LU Xuzhen(Outpatient Department,Jiaxing Center for Disease Control and Prevention,Jiaxing 314050,China;Infectious Diseases Department,the First Hospital of Jiaxing,Jiaxing 314001,China)

机构地区：[1]嘉兴市疾病预防控制中心门诊部,浙江嘉兴314050 [2]嘉兴市第一医院感染科,浙江嘉兴314001

出　　处：《健康研究》2024年第5期528-533,共6页Health Research

基　　金：嘉兴市科技计划项目(2023AY11040)。

摘　　要：目的 比较分析2017—2022年浙江省其他感染性腹泻病自回归移动平均(autoregressive integrated moving average, ARIMA)模型预测精度,探索更为精准的预测模型以指导传染病防控。方法 收集2011年4月—2022年12月的浙江省其他感染性腹泻病发病率资料,将2011年4月—2021年的数据分为6个时间段,拟合优选出各数据段的最优ARIMA模型,分别对2017—2022年浙江省其他感染性腹泻病的发病率进行预测,以对应年度实际发病率验证模型,比较各数据段最优模型的年平均相对误差和月相对误差。结果 2011年4月—2022年12月,浙江省总计报告其他感染性腹泻病1 272 546例,年均发病率为186.97/10万。拟合的6个最优ARIMA模型对2017—2022年预测值与实际值的年平均相对误差分别为19.27%、28.59%、12.46%、77.87%、16.53%、40.21%,最小的月相对误差分别为0.69%、1.16%、0.57%、3.27%、0.45%、8.07%。结论 虽然ARIMA模型预测其他感染性腹泻病有时会存在较大误差,但短期预测仍可以为该病的早期精准防控提供参考依据。Objective To compare and analyze the prediction accuracy of the autoregressive integrated moving average(ARIMA)model for other infectious diarrhea diseases in Zhejiang Province from 2017 to 2022,and to explore a more accurate predictive model to guide the prevention and control of infectious diseases.Methods To collect the incidence data of other infectious diarrhea diseases from April 2011 to December 2022 in Zhejiang Province,to segment the data and fit the optimal ARIMA model for each data segment.The incidence of other infectious diarrhea diseases in Zhejiang Province from 2017 to 2022 was predicted using these models,and the actual incidence of the corresponding years was used to validate the models.The annual average relative error and monthly relative error of the optimal models for each data segment were compared.Results From April 2011 to December 2022,a total of 1272546 cases of other infectious diarrhea diseases were reported in Zhejiang Province,with an average annual incidence rate of 186.97 per 100000 population.The annual average relative errors between the predicted and actual values for the six best-fitted ARIMA models from 2017 to 2022 were 19.27%,28.59%,12.46%,77.87%,16.53%,and 40.21%,respectively.The smallest monthly relative errors were 0.69%,1.16%,0.57%,3.27%,0.45%,and 8.07%,respectively.Conclusions Although the ARIMA model sometimes exhibits larger errors in predicting other infectious diarrhea diseases,it can still provide a reference for the early and precise prevention and control of the disease in the short term.

关 键 词：其他感染性腹泻病 ARIMA模型 预测精度 

分 类 号：R-183[医药卫生]