基于GRNN的组合预测模型在传染病发病率预测中的应用  

Application of a Combination Forecasting Model Based on GRNN for Incidence of Pulmonary Tuberculosis

作　　者：叶晓军[1] 沈毅[2] 任茹香[2,3] 范伟忠 

机构地区：[1]义乌市疾病预防控制中心,浙江义乌322000 [2]浙江大学公卫学院 [3]绍兴市卫生监督所

出　　处：《浙江预防医学》2012年第1期8-13,共6页Zhejiang Journal of Preventive Medicine

基　　金：浙江省科技厅赞助项目(2007C23001)

摘　　要：目的研究基于GRNN的组合预测模型拟合传染病发病率的优越性和不足。方法以浙中某市1998—2008年的肺结核发病率为研究资料,分别构建了灰色模型和ARIMA模型,以这两种模型为基础构建了基于GRNN的组合预测模型。结果残差修正GM(1,1)模型、ARIMA(1,0,1)*(1,1,0)12模型、基于GRNN的组合预测模型的MSE,MAE,MAPE和MER分别为37.451,5.692,53.69%,48.51%;18.509,3.761,35.13%,32.05%;9.961,2.571,25.6%,21.9%。结论基于GRNN的组合预测模型的预测精度优于两种单项模型。Objective To establish a combination forecasting model based on GRNN which could be used to predict the incidence rate of the infection diseases and evaluate its advantages and weakness. Methods Grey model and ARIMA model were established with the pulmonary tuberculosis incidence of the city in middle of Zhejiang province from 1998 to 2008. GRNN combination forecasting model was established based on the Grey and ARIMA model. Results MSE, MAE, MAPE and MER of GM （ 1, 1 ） model residual error correction, ARIMA （ 1, 0, 1 ） ＊ （ 1, 1, 0） 12, GRNN combination forecasting model were 37. 451,5. 692, 53. 69%, 48. 51% ; 18. 509, 3. 761, 35.13%, 32. 05 % ; 9. 961, 2. 571, 25.6%, 21.9%, respectively. Conclusion The forecast accuracy of the GRNN combination forecasting model was better than single GM model and single ARIMA model.

关 键 词：灰色模型 ARIMA模型 基于GRNN的组合预测模型 发病率预测 

分 类 号：R181.81[医药卫生—流行病学] R521[医药卫生—公共卫生与预防医学]