基于分数整合自回归移动平均模型的山西省手足口病预测研究  

Prediction of hand foot and mouth disease based on autoregressive fractionally integrated moving average model in Shanxi

作　　者：王晓瑞[1,2] 刘元[2] 王彤[1] Wang Xiaorui;Liu Yuan;Wang Tong(Department of Statistics,School of Public Health,Key Laboratory of Coal Environmental Pathogenesis and Prevention,Ministry of Education,China,Shanxi Provincial Key Laboratory of Diagnosis,Treatment,Prevention and Control of Major Infectious Diseases,Shanxi Medical University,Taiyuan 030001,Shanxi,China;Shanxi Provincial Center for Disease Control and Prevention,Taiyuan 030012,Shanxi,China)

机构地区：[1]山西医科大学公共卫生学院统计学教研室、煤炭环境致病与防治教育部重点实验室、重大传染性疾病诊治与防控山西省重点实验室,山西太原030001 [2]山西省疾病预防控制中心,山西太原030012

出　　处：《疾病监测》2023年第7期865-871,共7页Disease Surveillance

基　　金：山西省科技重大专项(No.202005D121008);山西省重点研发计划(No.202102130501003)。

摘　　要：目的探讨分数整合自回归移动平均(ARFIMA)模型在手足口病发病率预测中的可行性。方法基于Python语言,以山西省2008年1月至2021年8月手足口病发病率数据为训练集建立ARFIMA模型和自回归移动平均(ARIMA)模型,以2021年9月至2022年8月数据为测试集对所构建的两种模型进行效果评价,选用最优模型对2022年9月至2023年8月山西省手足口病发病率做出预测。结果建立ARFIMA(4,0.05,5)模型和ARIMA(5,1,2)模型,残差白噪声检验P≥0.05。利用构建好的ARFIMA模型和ARIMA模型对测试集进行预测,平均绝对误差分别为0.92、1.58,平均绝对百分比误差分别为1.28、1.67,均方误差分别为1.18、3.56,均方根误差分别为1.09、1.89。使用较优ARFIMA(4,0.05,5)模型预测2022年9月至2023年8月山西省手足口病月发病率在0.13/10万~3.51/10万。结论相比ARIMA模型,考虑了序列长记忆性的ARFIMA模型可较准确地预测山西省手足口病发病趋势,在手足口病防控中具有现实意义。Objective To explore the feasibility of autoregressive fractionally integrated moving average(ARFIMA)model in predicting the incidence of hand foot and mouth disease(HFMD).Methods Based on the Python language,the ARFIMA model and autoregressive integrated moving average(ARIMA)model were established by using the incidence data of HFMD in Shanxi province from January 2008 to August 2021 as the training set,and the data from September 2021 to August 2022were used as the test set to evaluate the effects of these two models constructed.The optimal model was selected to predict the incidence of HFMD in Shanxi from September 2022 to August 2023.Results ARFIMA(4,0.05,5)model and ARIMA(5,1,2)model were constructed,and the residual white noise test indicated that p-values were greater than 0.05.Using the constructed ARFIMA model and ARIMA model for prediction of the test set,the mean absolute errors were 0.92 and 1.58,the mean absolute percentage errors were 1.28 and 1.67,the mean square errors were 1.18 and 3.56,and the root mean square errors were 1.09 and 1.89,respectively.The optimal ARFIMA(4,0.05,5)model was used to predict the monthly incidence of HFMD in Shanxi from September 2022 to August 2023,indicating that the incidence of HFMD would range from 0.13/100000 to3.51/100000.Conclusion Compared with the ARIMA model,the ARFIMA model,which takes the long memory of the series into account,can predict the incidence trend of HFMD more accurately in Shanxi and is more feasible in the prevention and control of HFMD.

关 键 词：手足口病 长记忆性 分数整合自回归移动平均 PYTHON 模型预测 

分 类 号：R211[医药卫生—中医学] R183.1