ARIMA-BP神经网络组合预测模型在卫生人力资源配置中的应用  

Application of ARIMA-BP neural network combination forecast model in heath human resource allocation

作　　者：翟向明[1] 朱秋丽[2] 何晓敏[1] 

机构地区：[1]滨州医学院公共卫生与管理学院,山东省烟台市264003 [2]山东中医药高等专科学校

出　　处：《中国医院统计》2015年第6期430-434,共5页Chinese Journal of Hospital Statistics

基　　金：山东省统计科研重点课题(KT13165)

摘　　要：目的通过构建组合预测模型，提高预测的精确度，为相关部门制定卫生人力资源规划，合理配置卫生资源提供依据。方法将ARIMA和BP神经网络有机结合，建立ARIMA—BP神经网络模型，用ARIMA模型对卫生人力资源数据的线性部分建模预测，从线性模型中保留的残差将只保留非线性关系，然后用BP神经网络模型对非线性残差构成的序列进行预测，将两个模型的预测值相加作为该序列的预测值，BP神经网络模型预测产生的随机误差作为ARIMA-BP神经网络模型的预测误差，从而得到最后的预测结果。最后应用山东省1995--2012年卫生技术人员数量的数据对模型进行验证，并对2013--2015年山东省的卫生技术人员数量进行预测。结果通过山东省1995--2012年卫生技术人员数量的数据对模型验证，显示用ARIMA模型预测线性数据，用BP神经网络预测ARIMA的误差，形成组合预测模型，降低了ARIMA模型的预测误差，提高了预测的精度，同时通过组合模型预测2013年、2014年和2015年山东省卫生技术人员的数量分别为581297人、635013人和686465人。结论组合预测模型能够较大限度地利用各种预测样本信息，比单个预测模型考虑问题更系统全面，从而提高预测精度，可以在卫生入力资源预测中推广应用。Objective By constructing combination forecast models to improve the prediction accuracy, and to provide the basis for the rational allocation of health resources. Methods ARIMA and BP neural network were combined to establish an ARIMA-BP neural network model. ARIMA model is used to predict the linear part of the data of health human resource, and the residual of the model is nonlinear, and then the BP neural network model is used to forecast the nonlinear residual sequence. The prediction of the two models is added as the forecast value of the sequence, and the random error generated by the BP neural net- work model is used as the prediction error of the ARIMA-BP neural network model, and the final prediction result is obtained. Finally the data of the number of health technical personnel in Shandong Povince from 1995 to 2012 were applied to verify the model, and to forecast the number of health technical personnel in Shandong Province from 2013 to 2015. Results The verifi- cation of the number of health technical personnel data of Shandong Povince from 1995 to 2012 showed that the combination fore- cast model which was formed by using ARIMA model to predict the linear ＇data, and using BP neural network to predict the error of ARIMA, reduced the ARIMA model prediction error and improved the accuracy of prediction. The combined model forecast the numbers of health technicians in 2013, 2014 and 2015 in Shandong Province were 581 297,635 013 and 686 465 respective- ly. Conclusion -The combination forecasting model can make use of a variety of sample information to a large extent, and con- sider the problem more systematically than a single comprehensive predictive model to improve prediction accuracy. Therefore, it may be applied in human resources for health.

关 键 词：ARIMA BP神经网络 卫生人力资源 

分 类 号：R338[医药卫生—人体生理学]