自回归滑动平均混合模型模型在某三甲医院肺结核病门诊量预测中的应用  

Application of autoregressive moving average mixed model in the prediction of number of pulmonary tuberculosis outpatients in a Grade Ⅲ class A hospital

作　　者：边香 郭奇 侯晓芳 杨菁 郭柯宇 高永桂[4] 饶华祥 Bian Xiang;Guo Qi;Hou Xiaofang;Yang Jing;Guo Keyu;Gao Yong-gui;Rao Huaxiang(Department of Public Health and Preventive Medicine,Changzhi Medical College,Changzhi 046000,China;不详)

机构地区：[1]长治医学院公共卫生与预防医学系,046000 [2]长治医学院临床医学系,046000 [3]长治医学院麻醉学系,046000 [4]长治医学院附属和平医院预防保健科

出　　处：《实用医技杂志》2020年第12期1606-1608,共3页Journal of Practical Medical Techniques

基　　金：山西省大学生创新创业训练项目(2020391);长治医学院博士启动基金(BS201921)。

摘　　要：目的探讨自回归滑动平均混合模型(ARIMA)在肺结核病门诊量短期预测中的效果,为合理安排医院门诊资源配置及为肺结核病的防控工作提供理论依据。方法按月汇总2010—2019年晋东南地区某三甲医院的肺结核病就诊人数,利用2010—2018年的数据用传统建模法和专家建模器2种方法构建ARIMA模型,根据贝叶斯信息准则(BIC值)拟合最优模型,并进行拟合度分析,预测2019年肺结核门诊量并与实际值比较,综合评判其预测效果。结果采用传统建模法,依据参数选择原则,多次探索后确定最优模型为ARIMA(0,1,1)(0,1,1)12,而用专家建模器建立的ARIMA模型为ARIMA(0,0,1)(1,0,0)12,2个模型的平稳的R2分别为0.666、0.199,BIC值分别为5.843、5.689。最佳模型为ARIMA(0,1,1)(0,1,1)12。预测结果显示实际值绝大多数均落入95%预测区间内,预测效果较好。结论基于季节波动的ARIMA模型预测效果较理想,可以用于该院肺结核门诊量的短期预测。Objective To explore the effect of autoregressive moving average mixed model(ARIMA)on short-term prediction of tuberculosis outpatient numbers,and to provide theoretical basis for rational allocation of hospital resources and prevention and control of tuberculosis.Methods The number of tuberculosis outpatients were summarized monthly from 2010 to 2019 in a Grade Ⅲ class A hospital located in southeast of Shanxi province.ARIMA model were built by traditional modeling method and expert modeler respectively using the data from 2010 to 2018,which the optimal model was fitted based on bayesian information criterion(BIC),and performed the fitting degree analysis.Then the number of tuberculosis outpatients during 2019 was forecast using the optimal model,which prediction effect was evaluated by comparing with the actual number of tuberculosis outpatient.Results The optimal model was determined as ARIMA(0,0,1)(1,0,0)12 after multiple explorations by using the traditional modeling method according to the principle of parameter,and ARIMA model ARIMA(0,0,1)(1,0,0)12 established by the expert modeler respectively,which the stationary R2 was 0.666 and 0.199,and BIC value was 5.843 and 5.689.And ARIMA(0,0,1)(1,0,0)12 was chose as the optimal model,in which most of the actual values fall into the 95% prediction interval with the good prediction effect.Conclusion The prediction effect of ARIMA model based on seasonal fluctuation is ideal,which could be used for short-term prediction of number of tuberculosis outpatient in this hospital.

关 键 词：肺结核 ARIMA模型 门诊量 预测 

分 类 号：R521[医药卫生—内科学]