基于ARIMA乘积季节模型和Holt-Winters季节模型的梅毒月发病率预测
Application of seasonal model of ARIMA and Holt-Winters in prediction of the monthly incidence of syphilis

目的:探讨ARIMA乘积季节模型和Holt-Winters季节模型在我国梅毒月发病率预测中的应用价值。方法:以2005年1月至2015年12月梅毒月发病率数据为基础,运用SPSS 22.0和Eviews 8.0分别建立ARIMA乘积季节模型和Holt-Winters季节模型,采用2016年1至6月的实际数据验证模型,评价指标是预测误差和平均绝对误差(MAE)。选择精度较高模型预测2016年7至12月梅毒月发病率。结果:MAE的比较结果表明ARIMA乘积季节模型预测精度优于Holt-Winters季节模型,最优模型是ARIMA(1,1,1)×(0,1,1)12,模型口径为:(1-B)(1-B12)(1+0.374B)xt=(1+0.740B)(1+0.775B12)εt,2016年7至12月梅毒月发病率的预测结果(1/10万)分别为3.107、2.989、2.879、2.658、2.631、2.644。结论:ARIMA乘积季节模型具有较高的预测精度,可较好地拟合全国梅毒月发病率的演变趋势。
Aim: To explore the application value of ARIMA and Holt-Winters seasonal model for predicting the monthly incidence of syphilis.Methods: SPSS 22.0 and Eviews 8.0 were used to establish the seasonal model of ARIMA and Holt-Winters based on the data of the monthly incidence of syphilis in China from January 2005 to December 2015.Then the actual data from January to June in 2016 were used to confirm the predicted results. The prediction evaluation index was error and MAE. The data from July to December in 2016 were forcasted by the model with higher precision in the similar manner.Results: In the comparison of MAE, the prediction accuracy of the seasonal ARIMA model was higher than the Holt-Winters seasonal model. The optimal model for the monthly incidence was ARIMA(1,1,1)×(0,1,1)12, the model equation was(1-B)(1-B12)(1+0.374B)xt=(1+0.740B)(1+0.775B12)εt. The predicted results of the monthly incidence of syphilis(1/100 000)from July to December in 2016 were 3.107, 2.989, 2.879, 2.658, 2.631, 2.644.Conclusion: The seasonal ARIMA model features higher predictive accuracy, and could agree well with the trend of the monthly incidence of syphilis