基于ARIMA-BPNN模型对医院感染患病率的预测研究  

Prediction of prevalence rate of nosocomial infections based on ARIMA-BPNN model

作　　者：王高帅[1] 陈晓娟[2] 梁进娟[1] 

机构地区：[1]郑州大学第二附属医院感染管理科,河南郑州450001 [2]郑州大学第二附属医院护理部,河南郑州450001

出　　处：《中华医院感染学杂志》2017年第2期448-451,共4页Chinese Journal of Nosocomiology

基　　金：河南省医学科技攻关计划项目(201602113)

摘　　要：目的采用ARIMA模型和ARIMA-BPNN联合模型对某大型三甲医院医院感染患病率预测进行建模,预测医院感染患病率变化趋势,评价不同预测模型的预测效果。方法利用河南省某大型三甲医院2013年1月-2014年12月的医院感染患病率数据作为训练集,建立ARIMA模型和ARIMA-BPNN联合预测模型,选取2015年1-6月的患病率数据作为检验集,评价模型的预测效能。结果 ARIMA模型和ARIMA-BPNN联合模型的预测值的平均偏差及平均相对偏差分别为0.086,6.132%和0.015,1.080%;组合模型的预测效能优于ARIMA模型。结论 ARIMA-BPNN联合模型能有效模拟预测医院感染患病率,对预防和控制医院感染的发生具有重要的应用价值。OBJECTIVE To establish the models for prediction of prevalence rate of nosocomial infections in a largescale three A hospital by using ARIMA model and ARIMA-BPNN model, predict the changing trend of prevalence rate of nosocomial infections, and evaluate the prediction effects. METHODS The data of prevalence rate of nosoco- mial infections were collected from a large-scale three A hospital in Henan province from Jan 2013 to Dec 2014 and were used as the training set, the ARIMA model and ARIMA-BPNN combination prediction model were established, and the data of prevalence rate of nosocomial infections that were collected from Jan 2015 to Jun 2015 were used as the testing set. The predictive performances of the models were assessed. RESULTS The mean deviation of the predicted value of the ARIMA model was 0. 086, the ARIMA-BPNN combination model 0. 015; the mean relative deviation of the ARIMA model was 6. 132%, the ARIMA-BPNN combination model 1. 080%; the predictive performance of the combination model was superior to that of the ARIMA model. CONCLUSION The ARIMA- BPNN combination model can effectively predict the prevalence rate of the nosocomial infection, and it has significant value in prevention and control of the nosoeomial infections.

关 键 词：ARIMA模型 ARIMA-BPNN联合模型 医院感染 预测 患病率 

分 类 号：R181.32[医药卫生—流行病学]