文章摘要
张生奎,王镇德,杨 荔,张乐言,王永斌,袁聚祥.基于SARIMA⁃ERNN组合模型预测我国细菌性痢疾发病率[J].南京医科大学学报,2019,(6):925~931
基于SARIMA⁃ERNN组合模型预测我国细菌性痢疾发病率
Forecasting the incidence of bacillary dysentery in china based on SARIMA⁃ERNN combination model
投稿时间:2018-05-23  
DOI:10.7655/NYDXBNS20190628
中文关键词: 预测模型  细菌性痢疾  发病率
英文关键词: predictive model  bacterial dysentery  incidence
基金项目:河北省重点职业病防治技术研究 (13277709D)
作者单位
张生奎 华北理工大学公共卫生学院河北 唐山 063210 
王镇德 华北理工大学公共卫生学院河北 唐山 063210 
杨 荔 华北理工大学公共卫生学院河北 唐山 063210 
张乐言 华北理工大学公共卫生学院河北 唐山 063210 
王永斌 华北理工大学公共卫生学院河北 唐山 063210 
袁聚祥 华北理工大学公共卫生学院河北 唐山 063210 
摘要点击次数: 16
全文下载次数: 51
中文摘要:
      目的:探讨季节自回归滑动平均混合(SARIM)?Elman神经网络(ERNN)组合模型在我国细菌性痢疾发病率预测中的利用价值。方法:使用2005年1月—2016年12月我国细菌性痢疾的月发病率资料作为训练集分别建立SARIMA模型、ERNN模型以及SARIMA?ERNN组合模型,2017年我国细菌性痢疾的月发病率资料作为测试集用于模型间的效果评价。结果:SARIMA模型拟合及预测的平均相对误差(average relative error,MRE)、平均误差率(average error rate,MER)、均方根误差(root mean squared predict error,RMSE)和平均绝对误差(mean absolute error,MAE)分别为5.661 37、0.061 81、0.001 45、0.000 94和5.596 40、0.051 77、0.004 54、0.000 34;ERNN模型拟合及预测的MRE、MER、RMSE和MAE分别为5.348 57、0.056 05、0.017 08、0.000 79和5.544 30,0.044 55、0.000 36、0.00030;SARIMA?ERNN组合模型拟合及预测的MRE、MER、RMSE和MAE分别为4.942 52、0.047 33、0.001 15、0.000 72和4.251 30、0.044 19、0.000 38、0.000 29。结论:与ERNN模型和SARIMA模型相比,SARIMA?ERNN组合模型具有较高的有效性和合理性,可以用于我国细菌性痢疾的短期预测和早期预警。
英文摘要:
      Objective:To explore the application of SARIMA?ERNN combination model in predicting the incidence of bacterial dysentery in China. Methods:Using the monthly incidence data of bacterial dysentery in China from January 2005 to December 2016 as the training set,the SARIMA model,the ERNN model,and the SARIMA?ERNN combination model were established respectively. The monthly incidence of bacterial dysentery in China in 2017 which was used to test the efficacy of the above models was used as a test set. Results:The MRE,MER,RMSE and MAE fitted and forecasted by SARIMA model were 5.661 37,0.061 81,0.001 45,0.000 94 and 5.596 40,0.051 77,0.004 54,0.000 34 respectively. The MRE,MER,RMSE and MAE fitted and forecasted by ERNN model fit and predicted were 5.348 57,0.056 05,0.017 08,0.000 79 and 5.544 30,0.044 55,0.000 36,0.000 30 respectively;The MRE,MER,RMSE and MAE fitted and forecasted by SARIMA?ERNN combination model were 4.924 25,0.047 33,0.001 15,0.000 72 and 4.251 30,0.044 19,0.000 38,0.000 29 respectively. Conclusion:The SARIMA?ERNN combination model has high validity and reasonability that can be used for short?term forecasting and early warning of bacterial dysentery in China.
查看全文   查看/发表评论  下载PDF阅读器
关闭