季节性指数平滑模型在预测钦州市结核病发病人数中的应用

A pplication of Seasonal Exponential Smoothing Model in Predicting the Incidence of Tuberculosis in Qinzhou

目的:利用指数平滑模型对钦州市2010—2021年结核病月发病报告数据进行建模,探讨数学模型在结核病疫情预测中的适应性,为钦州市结核病的防控工作提供参考依据。方法:收集钦州市2010—2021年结核病月发病报告例数并建立月发病数据库,基于2010年1月—2020年12月的月发病报告人数建立指数平滑模型,比较2021年1—12月预测值和实际值对模型性能进行检验,并对2022年结核病发病情况进行预测。结果:2010—2021年,钦州市累计报告肺结核病例35572例,年均报告91.82/10万。钦州市肺结核病例报告数在2010—2014年呈现上升趋势,2015—2021年期间呈现逐年下降趋势。肺结核的月发病报告病例数在每年1—3月、10—12月出现高峰,4—9月处于较低发病水平。比较多种季节性指数平滑模型拟合结果,发现Winters加法模型预测效果最好。该模型Ljung-BoxQ检验的P值为0.31、StationaryR2和R2分别为0.639和0.835,RMSE、MAPE分别为23.782和7.52%,适合于本次研究。结论:2010—2021年广西钦州市肺结核疫情总体呈下降趋势,指数平滑模型拟合效果较好,可用于该市肺结核疫情的预测。

OBJECTIVE To build a model of exponential smoothing for modeling the monthly incidence of tuberculosis in Qin⁃zhou from 2010 to 2021 years,and to explore the adaptability of the mathematical model in the prediction of tuberculosis epidemic situation,and to provide reference for the prevention and control of tuberculosis in Qinzhou.METHODS The monthly number of tu⁃berculosis incidence cases in Qinzhou from 2010 to 2021 was collected and the monthly incidence database was established.The in⁃dex smoothing model was established based on the number of tuberculosis incidence cases reported from January 2010 to December 2020,and the model performance of the predictive value and actual value between January and December 2021 was compared with the test,and the incidence of tuberculosis in 2022 was predicted.RESULTS From 2010 to 2021,a total of 35572 tuberculosis cases were reported in Qinzhou,with 91.82/1010000 annually.The number of tuberculosis cases reported in Qinzhou city showed an up⁃ward trend from 2010 to 2014,and showed a downward trend year by year from 2015 to 2021.The number of monthly reported cas⁃es of tuberculosis peaked from January to March and October to December every year,and was at a low level of incidence from April to September.Comparing the results of various seasonal exponential smoothing model fits,it found that the Winters addition model predicted the best.The P-values of the Ljung-Box Q test of this model were 0.31,0.639 and 0.835 for Stationary R2 and R2,respec⁃tively,and 23.782 and 7.52%for RMSE and MAPE,respectively,which are suitable for this study.CONCLUSION From 2010 to 2021,the tuberculosis epidemic situation in Qinzhou,Guangxi province generally showed a downward trend,and the exponential smoothing model had a good fitting effect,which can be used for the prediction of the tuberculosis epidemic situation in the city.