SARIMA-GRNN组合模型和SARIMA模型在流行性腮腺炎发病率预测中的应用  

Application of SARIMA-GRNN combination model and SARIMA model to predicting the incidence rate of mumps

作　　者：刘天[1,2] 姚梦雷 黄继贵[1] 吴杨[3] 陈琦[3] 童叶青[3] 陈红缨[3] 梅芳盛[1] LIU Tian;YAO Meng-lei;HUANG Ji-gui;WU Yang;CHEN Qi;TONG Ye-qing;CHEN Hong-ying;MEI Fang-sheng(Jingzhou Municipal Center for Disease Control and Prevention,Jingzhou,Hubei 434000,China;Chinese Field Epidemiology Training Program,Beijing 100050,China;Hubei Provincial Center for Disease Control and Prevention,Wuhan,Hubei 430079,China)

机构地区：[1]荆州市疾病预防控制中心,湖北荆州434000 [2]中国现场流行病学培训项目,北京100050 [3]湖北省疾病预防控制中心,湖北武汉430079

出　　处：《实用预防医学》2021年第1期108-112,共5页Practical Preventive Medicine

基　　金：湖北省卫生计生委疾控专项(WJ2016JT-002)。

摘　　要：目的探讨SARIMA-GRNN组合模型和SARIMA模型在流行性腮腺炎发病率预测中的应用,并对他们的预测效果进行比较。方法选取2004—2016年上海市流行性腮腺炎逐月发病率资料,基于2004年1月—2016年6月的数据建立SARIMA模型。利用2004—2015年流行性腮腺炎的SARIMA模型拟合值与实际值、时间因子训练SARIMA-GRNN组合模型,并运用2016年1—6月数据进行验证,筛选模型最优平滑因子(spread)。采用2016年7—12月数据进行回代验证模型的外推预测效果。评价指标包括平均绝对误差百分比(MAPE)、平均误差率(MER)、均方误差(MSE)和平均绝对误差(MAE)。结果SARIMA(0,0,2)(0,1,1)12为最优SARIMA模型。SARIMA-GRNN组合模型spread值为0.0037。SARIMA模型、SARIMA-GRNN组合模型拟合的MAPE、MER、MSE和MAE依次分别为16.19%、15.18%、0.14、0.25;2.93%、2.28%、0.01、0.04。SARIMA模型、SARIMA-GRNN组合模型预测的MAPE、MER、MSE和MAE依次分别为17.40%、17.26%、0.03、0.16;15.24%、15.50%、0.02、0.14。结论SARIMA-GRNN组合模型拟合及预测效果均优于SARIMA模型,但预测精度有待进一步提高。Objective To explore the application of seasonal autoregressive integrated moving average-generalized regression neural network(SARIMA-GRNN)combination model and seasonal autoregressive integrated moving average(SARIMA)model to forecasting the incidence rate of mumps,and to compare the predicated effect between them.Methods Data regarding the monthly incidence rate of mumps in Shanghai from 2004 to 2016 were collected,and the SARIMA model was established based on the data from January 2004 to June 2016.The SARIMA model fitted and actual values and time factors for the monthly incidence of mumps in 2004-2015 were used to train the SARIMA-GRNN combination model.The data from January to June in 2016 were used for verification,and the model optimal smoothing factor(spread)was screened.The data about the period from July to December in 2016 were introduced back to the established models to evaluate their predictive capacity.The indexes of fitting efficiency and predictive capacity of the models included mean absolute error percentage(MAPE),mean error rate(MER),mean square error(MSE),and mean absolute error(MAE).Results The SARIMA(0,0,2)(0,1,1)12 model was the most appropriate SARIMA model.The spread value of SARIMA-GRNN combined model was 0.0037.In the fitting phase,the MAPE,MER,MSE and MAE fitted by the single SARIMA model were 16.19%,15.18%,0.14% and 0.25%,respectively,and those fitted by the SARIMA-GRNN combination model were 2.93%,2.28%0.01% and 0.04%,respectively.The MAPE,MER,MSE and MAE predicted by the single SARIMA model were 17.40%,17.26%,0.03 and 0.16,respectively,and those predicted by the SARIMA-GRNN combination model were 15.24%,15.50%,0.02 and 0.14,respectively.Conclusions The SARIMA-GRNN combination model is superior to the SARIMA model in fitting and prediction of monthly incidence of mumps,but the prediction accuracy needs to be further improved.

关 键 词：SARIMA GRNN 组合模型 流行性腮腺炎 

分 类 号：R512.1[医药卫生—内科学]