EEMD⁃SARIMA⁃ETS组合模型在河南省人布鲁氏菌病发病趋势预测中的应用

Value ofthe EEMD⁃SARIMA⁃ETS combined model for predicting the trends in human brucellosis incidence in Henan Province

目的探索集成经验模态分解(EEMD)⁃季节性差分自回归滑动平均(SARIMA)⁃基于状态空间的指数平滑(ETS)组合模型在人布鲁氏菌病发病趋势预测中的应用。方法采用周期分解法分解2005—2019年河南省人布鲁氏菌病的发病趋势和季节模式;采用EEMD方法分解序列包含的不同时间尺度本征模态分量(IMFs)和剩余分量,将数据划分为训练集(2005—2018年)和测试集(2019年),采用ETS和SARIMA模型分别模拟IMFs和剩余分量,构建EEMD⁃SARIMA⁃ETS组合模型,并将其预测准确性与单一SARIMA和ETS模型进行比较。结果2005—2015年河南省人布鲁氏菌病发病率呈上升趋势(APC=53.26%),2016—2019年呈下降趋势(APC=-18.65%),发病高峰在每年的3—6月。获得河南省人布鲁氏菌病发病的最优模型:ETS(A,AD,A)、SARIMA(2,1,3)(0,1,0)12,及由最佳ETS、SAR1MA模型构成的EEMD⁃SARIMA⁃ETS组合。EE⁃MD⁃SARIMA⁃ETS组合模型平均绝对百分比误差(10.36%)<ETS模型(16.20%)<SARIMA模型(27.31%),平均绝对误差、均方根误差、平均误差率和均方根百分比误差同样也小于单一的ETS和SARIMA模型。结论河南省人布鲁氏菌病是季节性疾病,EEMD⁃SARIMA⁃ETS组合模型能准确地预测其流行趋势,可为人布鲁氏菌病精准化动态防控提供参考。

Objective To investigate the value of the ensemble empirical mode decomposition(EEMD)⁃seasonal autoregressive integrated moving average(SARIMA)⁃exponential smoothing based on state space(ETS)combined model for predicting the trends in human brucellosis incidence in Henan Province.Methods The trends in and seasonal distribution of human brucellosis incidence were decomposed in Henan Province using the periodic decomposition method.All data were divided into the training(2005 to 2018)and testing datasets(2019)according to the intrinsic mode functions(IMFs)and residual functions at different time scales in the de⁃composition series of the EEMD method.The IMFs and residual functions were simulated with ETS and SARIMA models to create the EEMD⁃SARIMA⁃ETS combined model,and the accuracy of the EEMD⁃SARIMA⁃ETS combined model,SARIMA model and ETS model for the prediction of human brucellosis incidence was compared.Results The incidence of human brucellosis appeared a tendency to⁃wards a rise in Henan Province from 2005 to 2015[annual percent change(APC)=53.26%],and a tendency towards a decline in from 2016 to 2019(APC=-18.65%),and the human brucellosis incidence peaked during the period from March to June each year.The optimal models for predicting human brucellosis incidence in Henan Province included the ETS(A,AD,A)model,the SARIMA(2,1,3)(0,1,0)12 model and the EEMD⁃SARIMA⁃ETS combined model.The lowest average absolute percentage error was seen in the EEMD⁃SARIMA⁃ETS combined model(10.36%),followed by the ETS model(16.20%),and the highest was seen in the SARI⁃MA model(27.31%).In addition,the average absolute error,root mean square error,average error rate and root mean square percent⁃age error of the EEMD⁃SARIMA⁃ETS combined model were lower than those of the ETS and SARIMA models.Conclusion Human brucellosis is a seasonal disease in Henan Province.The EEMD⁃SARIMA⁃ETS combined model is accurate to predict the trends in hu⁃man brucellosis in Henan Province,which may provide insights into the precision dynamic control of human brucellosis.