The application of a SARIMA-GRNN combination model for predicting monthly syphilis in-cidence in Jiangsu Province
CHEN Haiyan
ZHOU Luojing
Abstract：Objective To construct a combination model integrating seasonal autoregressive integrated moving average(SARIMA)and generalized regression neural network(GRNN)to pro-vide a new methodological approach for predicting the incidence trend of syphilis in Jiangsu Prov-ince.Methods Monthly syphilis incidence data from Jiangsu Province from January 2005 to De-cember 2019 were used to establish SARIMA and SARIMA-GRNN combined models,with a com-parative analysis of their fitting accuracy and predictive performance.Results The optimal SARI-MA model parameters were SARIMA(0,1,1)(0,1,1)12,and the best smoothing parameter(spread)for the SARIMA-GRNN combined model was 0.11.In the prediction of syphilis inci-dence in 2019,the root mean square errors(RMSE)for the SARIMA model and the SARIMA-GRNN combination model were 0.336 and 0.287,respectively,while the mean absolute percent-age errors(MAPE)were 10.43％and 8.83％,respectively.And the mean absolute errors(MAE)were 0.278 and 0.245,respectively.Compared to the SARIMA model,the SARIMA-GRNN combination model reduced the RMSE by 14.58％,the MAPE by 15.33％,and the MAE by 11.87％,indicating that the SARIMA-GRNN combination model outperformed the SARIMA model in prediction accuracy.Conclusions The SARIMA-GRNN combination model is more pre-cise than the SARIMA model in fitting monthly syphilis incidence data in Jiangsu Province and has higher prediction accuracy,making it suitable for predicting syphilis incidence in this province.
Keywords：syphilisseasonal autoregressive integrated moving average modelgeneralized regression neural networkprediction
Publication Date：2025-11-28
Online Publishing Date：2025-12-29(First online date of this platform, not the publication date of the document)
Pages：7( 791-797 )