ARIMA模型在安徽省梅毒发病预测中的应用

Application of ARIMA model in the prediction of syphilis incidence in Anhui Province

目的探讨应用时间序列基于季节性的差分自回归移动平均模型(autoregressive integrated moving average,ARIMA)预测安徽省梅毒的发病人数,为梅毒早期预警防控工作提供参考。方法借助R 3.6.2软件以安徽省2004年1月至2016年12月的梅毒发病人数进行模型拟合,利用所得到的模型对2017年1-12月的发病情况进行预测,并按照预测值与实际观察值之间的差异评价其预测效果。结果安徽省梅毒发病数呈上升趋势,且有明显的周期性。ARIMA(1,1,1)(0,1,2)12为最优模型,AIC=-264.81,BIC=-249.99。残差序列Box-Pierce检验结果λ^2=1.444,P=0.963,λ=10.459,P=0.576,差异无统计学意义(P>0.05),表明为白噪声序列。模型精度效果评价MAE=0.06;RMSE=0.09;MAPE=1.00%,说明模型拟合效果好。2017年数据以检验模型外推效果,MAPE=6.09%,表明模型外推效果较好,且实际值均落在预测值的95%的置信区间,模型预测效果比较理想。结论ARIMA(1,1,1)(0,1,2)12模型能较好的拟合、预测安徽省梅毒发病人数,为梅毒预警和防控工作提供理论依据。

Objective To explore the application of time series autoregressive integrated moving average(ARIMA)based on seasonal difference to predict the number of syphilis cases in Anhui Province,and to provide a reference for early warning and control of syphilis.Methods Using R 3.6.2 software,the number of syphilis cases in Anhui Province from January 2004 to December 2016 was used for model fitting,and the resulting model was used to predict the incidence from January to December 2017.The difference between the predicted value and actual observed value was compared to evaluate the prediction effect of this model fitting.Results The incidence of syphilis in Anhui Province was on the rise with obvious periodicity.ARIMA(1,1,1)(0,1,2)12 was the optimal model,with the AIC being-264.81 and the BIC being-249.99.Box-Pierce test showed that λ^2 value was 1.444(P=0.963),10.459(P=0.576),and the difference was not statistically significant(P>0.05),indicating that the residual sequence was white noise.The model accuracy effect evaluation showed that the MAE was 0.06,the RMSE was 0.09,and the MAPE was 1.00%,indicating that the model fitting effect was good.The 2017 data was used to test the effect of the model extrapolation,and the results showed MAPE=6.09%,indicating that the model extrapolation effect was good.The actual value fell within 95%confidence interval of the predicted value,and the model prediction effect was relatively ideal.Conclusion The ARIMA(1,1,1)(0,1,2)12 model could better fit and predict the number of syphilis cases in Anhui Province,which may provide a theoretical basis for early warning,prevention and control of syphilis.