指数平滑法模型与ARIMA模型在北京市通州区肺结核流行趋势预测中的效果评价

Effect evaluation of exponential smoothing model and ARIMA model in the epidemic trend prediction of tuberculosis in Tongzhou District, Beijing

目的比较指数平滑法模型和自回归移动平均(autoregrissive integrated moving average,ARIMA)模型对北京市通州区肺结核流行趋势预测效果,为肺结核预测和防控提供依据。方法收集“中国疾病预防控制信息系统”2005年1月—2019年12月北京市通州区肺结核月发病率,应用2005年1月—2018年12月肺结核发病率构建指数平滑法模型和ARIMA模型,预测2019年1—12月肺结核发病率,并与实际值进行比较,评价2种模型的拟合和预测效果。结果指数平滑法确定简单季节模型为最优预测模型,Ljung-Box Q=15.265,P=0.505,残差序列为白噪声序列,标准贝叶斯信息准则(bayesian information criterion,BIC)值为0.115,2019年1—12月肺结核发病率预测的相对误差为0.45%~33.62%,平均为10.57%。ARIMA模型确定ARIMA(0,1,1)(0,1,1)_(12)为最优模型,Ljung-Box Q=17.156,P=0.376,残差序列为白噪声序列,BIC值为0.539,2019年1—12月肺结核发病率预测的相对误差为1.90%~37.57%,平均为16.39%。指数平滑法模型拟合效果的相对误差低于ARIMA模型。结论指数平滑法模型对北京市通州区肺结核发病率流行趋势预测效果优于ARIMA(0,1,1)(0,1,1)_(12)模型,可用于北京市通州区肺结核疫情预测和防控。

Objective To explore the effects of exponential smoothing model and autoregressive integrated moving average(ARIMA)model in the epidemic trend prediction of tuberculosis in Tongzhou District,Beijing,and to provide a basis for of tuberculosis prevention and control.Methods Monthly incidence data of tuberculosis in Tongzhou District from January 2005 to December 2018 in China information system for disease control and prevention were collected,and the incidence of tuberculosis was used to establish an exponential smoothing model and an ARIMA model to predict the monthly incidence data of tuberculosis from January to December 2019.The fitting and prediction effects of the two models were evaluated by comparing the real data.Results The simple seasonal model was determined as the optimal prediction model by exponential smoothing method,Ljung-Box Q=15.265,P=0.505,and the residual sequence was a white noise.The value of bayesian information criterion(BIC)is 0.115,and the relative error of the incidence rate of tuberculosis from January to December 2019 was 0.45%-33.62%,with an average of 10.57%.ARIMA(0,1,1)(0,1,1)_(12)was determined as the optimal model by ARIMA model,Ljung-Box Q=17.156,P=0.376,and the residual sequence was a white noise,and the value of BIC is 0.539.And the relative error of the incidence rate of tuberculosis from January to December 2019 was 1.90%-37.57%,with an average of 16.39%.The relative error of fitting effect of exponential smoothing model was lower than ARIMA model.Conclusion The fitted effectively of exponential smoothing model was better than ARIMA(0,1,1)(0,1,1)_(12)modelin predicting the epidemic trend of pulmonary tuberculosis incidence in Tongzhou District,Beijing,which can be used for the prediction and control of tuberculosis in Tongzhou District,Beijing.