基于ARIMA-GRNN组合模型预测手足口病发病率  

Forecasting incidence of hand-foot-mouth disease with ARIMA-GRNN hybrid model

作　　者：霍荣瑞 吕榜军[2] 于雪婷[1] 杨博[1] 黄高明[1] 

机构地区：[1]广西医科大学,广西南宁530021 [2]柳州市疾病预防控制中心,广西柳州545007

出　　处：《应用预防医学》2016年第6期481-485,共5页Applied Preventive Medicine

摘　　要：目的探讨ARIMA-GRNN组合模型在某市手足口病发病率预测中的应用,并比较其与ARIMA模型的预测效果。方法通过收集2010年1月至2015年12月广西某地级市手足口病发病率资料,用SPSS 20.0建立ARIMA模型,用MATLAB R2014a建立ARIMA-GRNN组合模型,并用2015年7月至12月数据对模型的预测效果进行评价。结果 ARIMA(1,0,0)(1,1,0)12模型拟合及预测的平均绝对误差分别为1.727、2.077,均方根误差分别为2.727、2.803;ARIMA-GRNN模型拟合及预测的平均绝对误差分别为1.012、0.467,均方根误差分别为1.453、0.587。结论 ARIMA-GRNN模型的拟合和预测效果均优于ARIMA模型,可用于手足口病月发病率短期预测。Objective To explore the application of ARIMA-GRNN hybrid model and ARIMA model in the prediction incidence of hand-food-mouth Disease（HFMD） in a city of Guangxi, China and compare the predicative effectiveness of them. Methods The data of incidence of HFMD from January 2010 to December 2015 was collected and SPSS 20.0 was used to construct ARIMA model. MATLAB R2014a was used to establish ARIMA-GRNN hybrid model, afterwards, the data of incidence from July to December in 2015 was used to evaluate the prediction effectiveness . Results The mean absolute error of ARIMA（1, 0, 0）（1, 1, 0）12 model in fitting and forecast were 1.727 and 2.077 respectively ; the root mean square prediction error of them were 2.727 and 2.803 respectively. The mean absolute error of ARIMA-GRNN model in fitting and forecast were 1.012 and 0.467respectively ;the root mean square prediction error of them were 1.453 and 0.587 respectively. Conclusions ARIMA-GRNN hybrid model has a better effectiveness than ARIMA model and can be used in prediction the incidence of hand-food-mouth disease in a short time.

关 键 词：手足口病 ARIMA-GRNN 时间序列分析 预测 

分 类 号：R512.5[医药卫生—内科学]