自回归求和移动平均模型对临床供血量的分析预测  

Estimation of clinical blood consumption by the autore-gressive integrated moving average model

作　　者：陈迎春[1] 王晓霞 徐晓庆 万浬科[1] 廖耘[1] 何炯[1] 张永鹏[1] 

机构地区：[1]成都市血液中心,四川成都610041 [2]西南财经大学

出　　处：《中国输血杂志》2017年第10期1174-1178,共5页Chinese Journal of Blood Transfusion

基　　金：成都市重大基金项目(2013013)

摘　　要：目的分析成都地区临床月供血量的规律,以此建立临床血液月供血量预测的时间序列ARIMA模型和乘积季节ARIMA模型,并动态进行模型的分析对比,为血液中心管理工作提供科学依据。方法收集2006年至2016年成都市血液中心临床血液月供血量,建立ARIMA模型和乘积季节ARIMA模型,预测2016年10-12月和2017年1-3月临床血液月供血量。对备选的模型进行拟合优度的比较,筛选出最优的模型,并对模型的相对误差进行评价。结果 ARIMA(0,1,1)模型预测2016年10-12月和2017年1-3月的相对误差为1.71%、-7.45%、-3.14%、-7.66%、-15.25%、-9.74%。而ARIMA(0,1,1)×(1,1,1)12模型相对误差为2.51%、-3.75%、-2.58%、-5.21%、-8.11%、-7.34%。结论乘积季节ARIMA模型能够较好的预测短期临床供血量,持续修正的乘积季节ARIMA模型能更好的预测下一季度临床血液月供血量。Objective This study aims to analyse the trend of the blood volume consumed for clinical purposes in Chengdu to generate a time series and a multiplicative seasonal mathematical model utilizing the autore-gressive integrated moving average（ ARIMA） for blood demand prediction purposes. The data would be valuable for improved blood management in blood centers. Methods Both models were established based on the blood volume consumed during 2006 to 2016 in Chengdu to predict the clinical blood demand from Oct to Dec in 2016 and Jan to Mar in 2017. The goodness of fit of potential models were compared for optimized modeling and the relative error of these models were evaluated as well. Results For the ARIMA（ 0,1,1） Models,the relative deviation values for Truth vs Prediction were 1. 71%、-7. 45%、-3. 14%、-7. 66%、-15. 25%、-9. 74% while the ARIMA（ 0,1,1） ×（ 1,1,1） 12 elative models yielded deviation values at 2. 51%、-3. 75%、-2. 58%、-5. 21%、-8. 11%、-7. 34%. Conclusion The ARIMA（ 0,1,1） ×（ 1,1,1） 12 model can be well used to simulate the incidence trend of clinical blood demand and provide the potential theory basis for blood donor recruitment.

关 键 词：ARIMA模型 乘积季节ARIMA模型 临床月供血量 相对误差 

分 类 号：R457.1[医药卫生—治疗学]