新冠疫情期间山西省流行性感冒的预测效果研究  

Study on the prediction effect of influenza during COVID-19,Shanxi

作　　者：赵执扬 翟梦梦 李国华[2] 高雪峰 王旭春 任浩 崔宇 乔宇超 任家辉 仇丽霞 ZHAO Zhi-yang;ZHAI Meng-meng;LI Guo-hua;GAO Xue-feng;WANG Xu-chun;REN Hao;CUI Yu;QIAO Yu-chao;REN Jia-hui;QIU Li-xia(Department of Health Statistics,Shanxi Medical University,Taiyuan,Shanxi 030001,China;不详)

机构地区：[1]山西医科大学公共卫生学院卫生统计学教研室,山西太原030001 [2]山西省疾病预防控制中心

出　　处：《现代预防医学》2023年第4期724-729,768,共7页Modern Preventive Medicine

基　　金：国家自然科学基金(81973155)。

摘　　要：目的为解决新型冠状病毒肺炎期间流感异常发病率而导致的流感建模预测困难,本研究建立了季节性自回归求和移动平均(seasonal autoregressive integrated moving average,SARIMA)、长短期记忆神经网络(long-Short Term Memory,LSTM)模型以及基于奇异谱分析(Singular spectrum analysis,SSA)的SSA-LSTM模型,为新冠疫情期间山西省流感的高精度预测提供有效的科学依据。方法收集2014年第14周至2021年第13周山西省流感共7年的周度流感监测数据,利用时间序列分解(Seasonal and Trend decomposition using Loess,STL)分析其季节特征。以2014年第14周至2020年第13周的流感样就诊比(the ratio of influenza-like illness,ILI%)作为训练集分别建立SARIMA、LSTM、SSA-LSTM模型,以2020年第14周至2021年第13周(新冠疫情期间)的ILI%作为测试集,比较三种模型的拟合、预测性能,确定最佳模型。结果受新冠疫情的影响,山西省流感自2020年初变化复杂,ILI%出现了明显的下降。SSA-LSTM的均方误差(Mean Squared Error,MSE)、平均绝对误差(Mean Absolute Error,MAE)、均方根误差(Root Mean Squared Error,RMSE)在拟合、预测结果中均达到了最小。结论SSA-LSTM模型对新冠肺炎疫情期间山西省异常下降的流感发病率进行预测更加精准,为流感有效防控提供一定的参考,也为疫情期间其他传染病异常发病率的预测提供了方法学上的参照。Objective To address the difficulties of influenza prediction during COVID-19,Seasonal Autoregressive Integrated Moving Average(SARIMA),Long-Short Term Memory(LSTM),and Singular Spectrum Analysis-Long-Short Term Memory(SSA-LSTM)were established,to provided an effective scientific basis for high-precision prediction of influenza in Shanxi Province during the COVID-19.Methods Seasonal and Trend decomposition using Loess(STL)was adopted to analyze the seasonal characteristics of the ratio of influenza-like illness(ILI%)in Shanxi Province,China,from the 14th week in 2014 to the 13th week in 2021.ILI%from the 14th week of 2014 to the 13th week of 2020 was used as the training set to establish the 3 models,and ILI%from the 14th week of 2020 to the 13th week of 2021(during the COVID-19)was used as the test set to compare the fitting and prediction performance of the 3 models.Results Due to the influence of COVID-19,influenza in Shanxi has been complicated since 2020,and ILI%decreased significantly.MSE,MAE,and RMSE of the SSA-LSTM reached the minimum in fitting and prediction.Conclusion The SSA-LSTM is more accurate in predicting the abnormal decline of influenza in Shanxi Province during COVID-19,providing a leg-up for public policy,as well as a methodological reference for predicting the abnormal incidence of other infectious diseases during the epidemic.

关 键 词：流感 新型冠状病毒肺炎 SARIMA模型 LSTM模型 SSA-LSTM模型 预测 

分 类 号：R511.7[医药卫生—内科学] R181.81[医药卫生—临床医学]