辽宁省手足口病疫情季节性ARIMA模型预测效果评价

目的 利用时间序列法研究手足口病的发病趋势和流行特征，建立ARIMA乘积季节模型，对2017年7月-2018年12月辽宁省手足口病的月发病疫情情况及流行强度进行预测，为手足口病的预防监测工作提供依据。方法 收集辽宁省手足口病2012年1月-2016年12月的月发病报告数，采用Excel 2010建立辽宁省手足口病月报告发病数数据库，应用SPSS 23.0软件进行SARIMA模型的构建，拟合发病情况，对辽宁省手足口病2017年7月-2018年12月发病数进行预测，评价预测效果。结果 辽宁省手足口病发病特征以年为流行周期，季节性周期为12个月（s=12）。每年6-9月为该病的发病高峰期。最佳模型为SARIMA（0，1，0）×（1，1，0）₁₂季节性模型，模型残差Ljung-Box Q=18.564，P=0.354，序列为白噪声。预测平均相对误差为0.229，预测效果较好。结论 季节性ARIMA模型能较好的拟合辽宁省手足口病的发病流行趋势，能够比较直观准确的反映辽宁省手足口病的疫情发展情况，该模型适用于辽宁省手足口病的短期流行趋势的预测。

Prediction of hand-foot-mouth disease epidemic with seasonal autoregressive integrated moving average model

Objective To describe the prevalence and incidence trend of hand-foot-mouth disease (HFMD) using time series method and to establish a seasonal autoregressive integrated moving average (SARIMA) model for the prediction of HFMD epidemic from July 2017 to December 2018 in Liaoning province.Methods A prediction model was set up based on reported monthly data of HFMD from January 2012 through 2016 in Liaoning province using Excel and SPSS 23.0 software.Then the model was used to predict monthly incidents of HFMD from July 2017 to December 2018 and the prediction efficiency was evaluated.Results Yearly cycle and 12-month seasonal period of HFMD epidemic were observed.There was an incidence peak for HFMD epidemic from June to September during a year in Liaoning pro-vince.The established optimum model was SARIMA(0,1,0)×(1,1,0)₁₂.The Ljung-Box Q statistics is 18.564(P=0.354).The noise of residual series of the ARIMA model was white noise and the mean error was 0.229.Conclusion The established seasonal ARIMA model could well fit the monthly data of HFMD epidemic in Liaoning province and the model could be applied to predict short-term incident trend of HFMD in the province.