不同时间序列模型在潍坊市肾综合征出血热预测应用中的比较研究  

Comparison of Different Time Series Models in the Prediction of Hemorrhagic Fever with Renal Syndrome in Weifang

作　　者：郑良 高琦[1] 于胜男 石圆 孙明浩 王志强[2] 李秀君[1] Zheng Liang;Gao Qi;Yu Shengnan(Department of Epidemiology and Health Statistics,School of Public Health,Shandong University(250012),Jinan)

机构地区：[1]山东大学公共卫生学院流行病与卫生统计学系,250012 [2]山东省疾病预防控制中心传染病防治所

出　　处：《中国卫生统计》2024年第3期393-397,共5页Chinese Journal of Health Statistics

基　　金：国家重点研发计划项目(2019YFC1200500,2019YFC1200502)。

摘　　要：目的比较季节自回归移动平均模型(SARIMA)、长短期记忆网络(LSTM)、经验动态建模(EDM)在包含及不包含气象因素的情况下预测潍坊市肾综合征出血热(HFRS)发病的效果,探索最佳预测模型。方法选取2011年1月至2017年12月潍坊市HFRS月发病率分别构建SARIMA模型、单变量LSTM模型、单变量EDM模型,以及包含气象因素的SARIMAX模型、多变量LSTM模型、多变量EDM模型,对2018年1月至2018年12月的月发病率进行预测,并比较各模型的预测效果。结果SARIMA模型的平均绝对误差百分比(MAPE)为42.17%,SARIMAX模型未通过参数检验;单变量LSTM模型、多变量LSTM模型的MAPE分别为48.40%,16.19%;单变量EDM,多变量EDM模型的MAPE分别为55.00%,51.79%。结论包含气象因素的多变量LSTM模型对潍坊市HFRS发病率预测效果最佳,预测结果可为未来HFRS的防控提供参考。Objective To compare the effects of SARIMA,LSTM and EDM in predicting the incidence of HFRS in Weifang under different circumstances,and explore the best prediction model.Methods The monthly incidence of HFRS in Weifang from January 2011 to December 2017 was selected to construct the SARIMA model,univariate LSTM model,univariate EDM model,and SARIMAX model,multivariate LSTM model,and multivariate EDM model including meteorological factors.The monthly incidence from January 2018 to December 2018 was predicted,and the prediction effects of each model were compared.Results The MAPE of SARIMA model,univariate LSTM model,multivariate LSTM model,univariate EDM model,multivariate EDM model were 42.17%,48.40%,16.19%,55.00%,51.79%,respectively.Conclusion The multivariate LSTM model including meteorological factors had a good prediction effect on the incidence of HFRS in Weifang,and the prediction results could provide reference for the prevention and control of HFRS.

关 键 词：肾综合征出血热 气象因素 SARIMA模型 LSTM模型 EDM模型 

分 类 号：R195.1[医药卫生—卫生统计学]