基于EMD-Elman组合模型的细菌性痢疾发病率预测研究

Prediction of the incidence of bacillary dysentery based on the combined EMD-Elman model

目的为了解重庆市2009—2015年细菌性痢疾的流行特征,提高细菌性痢疾发病预测的准确率,比较经验模态分解(empirical mode decomposition,EMD)-Elman组合预测模型、Elman神经网络和季节自回归滑动平均模型(seasonal autoregressiove integrated moving average,SARIMA)模型在细菌性痢疾发病率预测的应用效果。方法整理分析2009—2014年重庆市细菌性痢疾报告病例流行特征。采用EMD方法对2009年1月—2015年5月重庆市报告的细菌性痢疾发病数据进行分解,对得到的一系列分量分别构建Elman神经网络并进行预测,并与Elman神经网络、SARIMA模型相比较。模型预测效果评价指标采用平均绝对误差(mean absolute error,MAE)、平均绝对百分比误差(mean absolute percentage error,MAPE)和均方根误差(root mean square error,RMSE)。结果2009-2014年重庆市细菌性痢疾平均年报告发病率为29.78/10万。发病高峰集中在每年6—10月,发病地区主要集中在主城区。EMD-Elman组合预测模型各评价指标均为最优,其RMSE、MAPE、MAE值分别为0.046、0.024、0.040;其次为Elman神经网络,其RMSE、MAPE、MAE值分别为0.093、0.051、0.081;SARIMA模型效果最差,其RMSE、MAPE、MAE值分别为0.151、0.070、0.117。结论EMD-Elman组合预测模型提高了细菌性痢疾发病率预测精度,是一种可靠的传染病时序预测方法,可为传染病的风险控制和疾病预防政策的提出提供一定的科学指导。

Objective To understand the epidemiological characteristics of bacillary dysentery in Chongqing from 2009 to 2015,to compare the application efficacies of combined EMD-Elman model,the Elman neural network and the SARIMA model in the prediction of incidence of bacillary dysentery in order to improve the accuracy of disease prediction.Methods Epidemiological data of reported cases with bacillary dysentery in Chongqing from 2009 to 2014 were retrieved and analyzed,and an incidence map was generated using ArcGIS 10.6 software.The incidence data of bacillary dysentery reported from January 2009 to May 2015 was decomposed by EMD and Elman neural network was constructed based on a series of components obtained to predict the incidence.The mean absolute error(MAE),mean absolute percentage error(MAPE)and root mean square error(RMSE)were adopted for the assessment of predictive efficacy of combined EMD-Elman model.Results The average annual reported incidence of bacillary dysentery in Chongqing from 2009 to 2014 was 29.78/100000,the peak season was from June to October and most cases were in urban areas.The combined EMD-Elman model was found to be the most accurate predictive model,with RMSE,MAPE and MAE values of 0.046,0.024 and 0.040,while the SARIMA model was the worst one,with RMSE,MAPE and MAE values of 0.151,0.070 and 0.117,respectively.Conclusions The combined EMD-Elman model improves the accuracy of predicting incidence of bacillary dysentery.It is a reliable time-series prediction method for infectious diseases,which may provide scientific guidance for the risk control of contagious diseases and the proposal of disease prevention policies.