应用ARIMA-GRNN模型对肾综合征出血热发病率时间序列数据的预测研究  

Application of ARIMA-GRNN Model to Predict the Incidence of Hemorrhagic Fever with Renal Syndrome

作　　者：吴伟[1] 郭军巧[2] 安淑一[2] 关鹏[1] 周宝森[1] 

机构地区：[1]中国医科大学公共卫生学院,110122 [2]辽宁省疾病预防控制中心

出　　处：《中国卫生统计》2015年第2期211-213,共3页Chinese Journal of Health Statistics

基　　金：国家自然科学基金项目(No.81202254);国家自然科学基金项目(No.30771860)

摘　　要：目的阐述ARIMA-GRNN模型预测肾综合征出血热发病率的方法和步骤,探讨其在综合征出血热发病率预测中的应用。方法利用辽宁省1962-2008年的肾综合征出血热发病率时间序列数据作为训练集,建立ARIMA模型和ARIMA-GRNN模型,选取2009-2011年的数据作为检验集,评价模型的拟合和预测效果。结果 ARIMA(2,1,1)模型和ARIMA-GRNN模型拟合值的平均误差绝对值分别为1.14和0.77;预测值的平均误差绝对值分别为0.53和0.20。ARIMA-GRNN模型的拟合和预测效果均优于ARIMA模型。结论 ARIMA-GRNN模型能有效模拟、预测肾综合征出血热的发病疫情,具有较强的推广应用价值。ABSTRACT： To describe the procedure of building ARIMA-GRNN model, and explore the value of potential application of the above model to predict the incidence of hemorrhagic fever with renal syndrome. ARIMA and ARIMA-GRNN models were established using the incidence of hemorrhagic fever with renal syndrome in Liaoning province from 1962 to 2008, and the fitting and predictive performances were evaluated using the data from 2009 to 2011. The mean absolute value of error for fitted value of the optimal ARIMA model and ARIMA-GRNN model were 1.14 and 0.77, respectively; and the mean absolute value of error for predicted value of the optimal ARIMA model and ARIMA-GRNN model were 1.14 and 0.77, respectively. This suggested that that the fitting and predictive efficacies of ARIMA-GRNN model were better than ARIMA model. ARIMA-GRNN model could effectively fit and predict the incidence of hemorrhagic fever with renal syndrome, which was of great application value for the prevention and control of hemorrhagic fever with renal syndrome.

关 键 词：肾综合征出血热 时间序列 发病率 预测 

分 类 号：R512.8[医药卫生—内科学]