SARIMA模型和BPNN模型在江苏省肺结核发病预测中的应用

Application of SARIMA model and BPNN model in the prediction of tuberculosis incidence in Jiangsu Province

目的探讨季节性差分自回归滑动平均(seasonal auto regressive integrated moving average,SARIMA)模型与反向传播神经网络(back propagation neural network,BPNN)模型在江苏省结核病发病数中的应用,评价两种模型的精确性,为江苏省制定肺结核防控策略提供参考。方法本研究以江苏省2011—2020年肺结核发病数据分别建立SARIMA模型和BPNN模型,以2021年1月—2022年6月的实际肺结核发病数进行检验,比较两种模型的预测精度和建模效果。结果SARIMA模型的均方根误差(root mean squared error,RMSE)、平均绝对误差(mean absolute error,MAE)、平均绝对百分比误差(mean absolute percentage error,MAPE)、总体相对误差值以及最小误差率分别为192、128、5.18%、0.05%、1.14%,BPNN模型的RMSE、MAE、MAPE、总体相对误差值以及最小误差率分别为301、246、11.03%、2.79%、1.44%,均高于SARIMA模型。SARIMA模型的拟合值和预测值明显更接近于实际值,预测效果更好。结论SARIMA能较好地拟合和预测江苏省肺结核病每月的发病率,可为该病的监测和防控工作提供依据。

Objective To investigate the application of the seasonal differential autoregressive moving average(SARIMA)model and back propagation neural network(BPNN)model in the incidence of tuberculosis in Jiangsu Province and to compare and evaluate the accuracy of the two models to provide a reference for the development of tuberculosis prevention and control strategies in Jiangsu Province.Methods The SARIMA and BPNN models were established by using the tuberculosis incidence data in Jiangsu Province from 2011 to 2020.The actual tuberculosis incidence from January 2021 to June 2022 was tested,and the prediction accuracy and modeling effect of the two models were compared.Results The root mean squared error(RMSE),mean absolute error(MAE),mean absolute percentage error(MAPE),overall relative error value,and minimum error rate of the SARIMA model were 192,128,5.18%,0.05%,and 1.14%.Moreover,the RMSE,MAE,MAPE,overall relative error value,and the minimum error rate of the BPNN model were 301,246,11.03%,2.79%,and 1.44%.Conclusions The SARIMA model can fit and predict the monthly incidence rate of pulmonary tuberculosis in Jiangsu Province quite well,providing a basis for monitoring and controlling the disease.