ARIMA模型与Holt-Winters指数平滑模型在武汉市流感样病例预测中的应用  

ARIMA model and Holt-Winters exponential smoothing method to predict influenza-like cases,Wuhan

作　　者：汪鹏[1] 彭颖[1] 杨小兵[1] 

机构地区：[1]武汉市疾病预防控制中心,湖北武汉430015

出　　处：《现代预防医学》2018年第3期385-389,共5页Modern Preventive Medicine

摘　　要：目的探讨比较ARIMA模型和Holt-Winters模型在武汉市流感样病例预测中的应用,为流感防控提供科学依据。方法利用武汉市2012年1月-2017年6月每周流感样病例比例数据拟合建立ARIMA模型和Holt-Winters指数平滑模型,预测2017年7-8月周流感样病例比例,并与实际流感样病例比例进行比较。结果 ARIMA最优模型为ARIMA(1,0,1)×(0,1,1)_(52),预测的平均相对误差为6.88%,Holt-Winters的最优模型为乘法模型,预测平均相对误差为13.79%。结论 ARIMA(1,0,1)×(0,1,1)_(52)模型拟合效果较好,预测精度更高,可用于武汉市流感样病例的预测。Objective To compare the efficiency of ARIMA model and Holt - Winters exponential smoothing method to predict the weekly proportion of influenza -like illness（ ILI% ）in Wuhan city, so as to provide a scientific reference for preventing and controlling influenza. Methods Weekly ILI% data in Wuhan city between January,2012 and June,2017 were collected to fit ARIMA model and Holt - Winters exponential smoothing model and to predict ILI% from July to August,2017 ,comparing to the true weekly ILI% data in the same period. Results The optimal ARIMA model was ARIMA （ 1,0,1 ） ＊ （ 0,1,1 ） 52, whose average relative error of prediction value was 6.88%. Holt - Winters muhiplicative model was the optimal model in exponential smoothing model with an average relative error of prediction value of 13.79%. Conclusion ARIMA （ 1,0,1 ） ＊ （ 0,1,1 ） 52 with a better fitting effect and a higher forecast precision can be used for predicting the weekly ILI% in Wuhan city.

关 键 词：流感样病例 ARIMA模型 指数平滑模型 预测 

分 类 号：R195[医药卫生—卫生统计学]