中国肺结核发病趋势的ARIMA乘积季节模型构建

Establishment of tuberculosis incidence trend model in China by multiple seasonal ARIMA model

目的根据我国肺结核发病趋势和流行病学特征,建立ARIMA乘积季节模型并进行预测。方法采用SPSS 25.0软件,对2004—2018年肺结核发病月度数据建模,预测2019年各月发病数,并与实际数据对比,分析预测效果。结果构建的ARIMA(2,1,0)(2,1,1)12模型的拟合系数R^(2)=0.762,拟合度较高,残差序列的白噪声通过检验。预测结果的平均绝对误差百分比(MAPE)为7.17%,且在3月、10月关键节点处,预测误差仅为4.61%、2.5%。结论构建的ARIMA乘积季节模型拟合及预测效果良好,能较好地描述我国肺结核的发病趋势,为防控物资储备和人员安排提供数据支持。

Objective To establish a multiple seasonal ARIMA model for predicting the infectious incidence of pulmonary tuberculosis in China according to the prevalence trend and epidemic characteristics of pulmonary tuberculosis. Methods SPSS 25.0 software was used to model the monthly pulmonary tuberculosis incidence in China from 2004 to 2018.The monthly incidence in 2019 was predicted and compared with the actual data to analyze the prediction effect. Results The fitting coefficient R^(2) of ARIMA(2,1,0)(2,1,1)12 model was 0.762,and the fitting degree of the model was high.The white noise of the residual sequence passed the test.The root mean square error(MAPE) of the prediction results was 7.17%,and the prediction errors at the key nodes in Mar and Oct were only 4.61% and 2.5%,respectively. Conclusion The fitting and prediction effect of established multiple seasonal ARIMA model are good, which can describe the incidence trend of tuberculosis in China and provide data support for material reservation and personnel arrangement for specific prevention and control measures.