指数平滑法在某院泌尿外科出院量及手术量预测中的应用  

Application of Exponential Smoothing Method in Predicting the Number of Discharges and Operations in the Department of Urology of a Hospital

作　　者：游晓平 吕庆文[1] 邹志武[1] 谢薇[1] 刘君[1] 梁彩霞 张巍[1] You Xiaoping;Lv Qingwen;Zou Zhiwu;Xie Wei;Liu Jun;Liang Caixia;Zhang Wei(Information Department,Zhujiang Hospital of Southern Medical University,Guangzhou 510000,Guangdong Province,China;不详)

机构地区：[1]南方医科大学珠江医院信息科,广东省广州市510000

出　　处：《中国病案》2021年第8期35-37,43,共4页Chinese Medical Record

摘　　要：目的基于指数平滑法,对某院泌尿外科出院量及手术量进行拟合评价及预测。方法提取某院泌尿外科2016年1月1日-2019年12月31日的出院量及手术量,其中2016年1月-2019年6月的出院量及手术量分别用于建立模型,采用指数平滑法对其进行拟合及评价。统计软件为SPSS20.0。对2019年7~12月的数据进行短期预测,比较实际值与预测值间的差异,评价其预测效果。结果(1)出院量:整体呈上升趋势,且有明显的季节特征。统计模型为季节性指数平滑模型中的Winters可加性模型。该模型平稳的R方为0.712,MAPE(Mean Absolute Percentage Error)为7.383%,Normalized BIC为7.579,Ljung-Box检验无统计学意义(Q 18=21.885,P>0.05),可认为该残差序列为白噪声序列。残差的ACF和PACF图显示为平稳序列。模型参数估计中平滑参数Alpha(Level)检验结果显示有统计学意义(t=2.620,P<0.05),估计值为0.308。各月实际值与预测值间的平均相对误差绝对值为6.28%。(2)手术量:序列有明显的季节特征。季节性指数平滑模型中的简单季节为最优模型。平稳的R方为0.689,MAPE为11.198%,Normalized BIC为8.395%,Ljung-Box检验无统计学意义(Q 18=12.500,P>0.05)。模型参数估计中平滑参数Alpha(Level)估计值为0.500,参数检验结果显示有统计学意义(t=3.434,P<0.05)。各月实际值与预测值间的平均相对误差绝对值为8.37%。结论指数平滑法适用于某院泌尿外科出院量及手术量的预测,可为该科医疗资源管理提供科学依据。Objective Based on the exponential smoothing method,the number of urology discharge and the number of operations in a hospital were fitted,evaluated and predicted.Methods The number of discharges and operations of the urology department of a hospital from January 1,2016 to December 31,2019 were extracted,in which the number of discharge and operations from January 2016 to June 2019 were used to establish models,and exponential smoothing method was used to fit and evaluate them.The statistical software was SPSS20.0.Make a short-term prediction of the data from July to December 2019,compare the difference between the actual value and the predicted value,and evaluate the prediction effect.Results(1)The number of discharged patients showed an upward trend as a whole,and had obvious seasonal characteristics.The statistical model is the Winters additive model in the seasonal exponential smoothing model.The R square of the model is 0.712,the MAPE(Mean Absolute Percentage Error)is 7.383%,the normalized BIC is 7.579,and the LjungmurBox test has no statistical significance(Q18 error 21.885,P>0.05).It can be considered that the residual sequence is a white noise sequence.The ACF and PACF diagrams of the residuals are shown as stationary series.In the parameter estimation of the model,the results of Alpha(Level)test showed that the smoothing parameters were statistically significant(t=2.620,P<0.05),and the estimated value was 0.308.The absolute value of the average relative error between the actual value and the predicted value of each month is 6.28%.(2)Number of operations:the sequence has obvious seasonal characteristics.The simple season in the seasonal exponential smoothing model is the optimal model.The stable R,MAPE and Normalized BIC were 0.689,11.198%and 8.395%,respectively,and the LjungmurBox test was not statistically significant(Q18=12.500,P>0.05).In the parameter estimation of the model,the estimated value of smoothing parameter Alpha(Level)was 0.500,and the result of parameter test showed that it was statistical

关 键 词：指数平滑法 出院量 手术量 预测 

分 类 号：R197.323[医药卫生—卫生事业管理]