ARIMA模型和指数平滑法在我国北方流感样病例预测中的应用和比较

Application and comparison of ARIMA model and exponential smoothing method to the prediction of influenza-like illness cases in northern China

目的比较ARIMA模型和指数平滑法对我国北方流感样病例的预测效果,为流感防控提供科学依据。方法利用我国北方2012年第1周—2018年第17周的每周流感样病例比例数据拟合建立ARIMA模型和指数平滑模型,预测2018年第18周—2019年第17周的流感样病例比例,对预测值与实际值进行比较。结果 ARIMA最优模型为ARIMA(0,1,1)(2,1,0)52,预测的均方根误差(root mean square error,RMSE)、平均绝对百分比误差(mean absolute percentage error,MAPE)和平均绝对误差(mean absolute error,MAE)分别为0.57%、8.98%、0.34%;指数平滑法的最优模型为简单季节性模型,预测的均方根误差(RMSE)、平均绝对百分比误差(MAPE)和平均绝对误差(MAE)分别为0.83%、15.24%、0.55%。结论 ARIMA(0,1,1)(2,1,0)52模型预测精度更高,可用于我国北方流感样病例的短期预测。

Objective To compare the efficiency of autoregressive integrated moving average(ARIMA)model and exponential smoothing method to predict cases of influenza-like illness(ILI%)in northern China,and to provide a scientific basis for preventing and controlling influenza.Methods Data concerning the weekly percentage of outpatient visits for influenza-like illness(ILI)in northern China between the first week of 2012 and the seventeenth week of 2018 were collected to fit and establish ARIMA model and exponential smoothing model.The weekly ILI%from the eighteenth week of 2018 to the seventeenth week of 2019 was predicted,and then the predictive value was compared with the actual value.Results The optimal ARIMA model was ARIMA(0,1,1)(2,1,0)52,and root mean square error(RMSE),mean absolute percentage error(MAPE)and mean absolute error(MAE)of the predictive value were 0.57%,8.98%and 0.34%,respectively.Simple seasonal model was the optimal model based on exponential smoothing method,and RMSE,MAPE and MAE of the predictive value were 0.83%,15.24%and 0.55%,respectively.Conclusions ARIMA(0,1,1)(2,1,0)52 has a higher forecast precision,and it can be used for short-term prediction of ILI cases in northern China.