摘要
目的为了弥补单一模型在医院门诊量预测中的不足,构建Arima-Elman耦合模型并探讨其合理性。方法基于R和Matlab,采用北京市某三甲综合医院2010年1月1日-2018年12月31日门诊量数据,分别构建Arima乘积季节模型、Elman神经网络和Arima-Elman耦合模型对该院2019年1月1日-2019年12月31日各月门诊量进行预测,并将预测值与真实值进行比较分析。结果Arima乘积季节模型构建的最佳模型为Arima(2,1,0)(0,1,1)12,预测的RMSE、MAE、MAPE分别为10546.18、8175.89、0.0420;Elman神经网络确定的最优隐含层神经元为10,预测的各项误差分别为9843.32、8105.12、0.0425;基于2种单一模型,分别以0.47和0.53为权重构建Arima-Elman耦合模型,预测的各项误差分别为8879.38、7089.15、0.0369。Arima-Elman耦合模型各项预测误差均明显小于2种单一模型。结论Arima-Elman耦合模型预测结果精度较高,对于医院门诊量的预测具有良好的适用性。
Objectives To make up for the shortcomings of single model in the prediction of outpatients in hospitals,the Arima-Elman coupling model was constructed and its rationality was discussed.Methods Based on R and Matlab,the monthly outpatient data of a Three A and Tertiary Hospital in Beijing from January 1 st,2010 to December 31 st,2018 were used to build the Arima product seasonal model,Elman neural network and their coupling model to predict the monthly outpatients of the hospital from January 1 st to December 31 st,2019,and then compared the predicted value with the true value.Results The best model constructed by the Arima product seasonal model was Arima(2,1,0)(0,1,1)12,and the predicted RMSE,MAE,and MAPE are 10546.18,8175.89,and 0.0420,respectively;the optimal hidden layer neuron determined by Elman neural network was 10,and the prediction errors were 9843.32,8105.12,0.0425,respectively;based on two single models,the Arima-Elman coupling model was constructed with weights of 0.47 and 0.53,respectively and the errors were 8887.38,7089.15,and 0.0369,respectively.The prediction errors of the Arima-Elman coupling model were significantly smaller than the two single models.Conclusions The prediction result of the Arima-Elman coupling model had a higher precision,and it had a good applicability for the prediction of hospital outpatients.
作者
李磊
孙瑞华
王蜜源
Li Lei;Sun Ruihua;Wang Miyuan(College of management,Beijing University of Chinese Medicine,Beijing 100029,China;不详)
出处
《中国病案》
2020年第11期68-72,共5页
Chinese Medical Record
作者简介
通信作者:孙瑞华。
