2008-2015年宁波市流感样病例预测模型分析

目的：建立宁波市流感样病例（ILI）的预测模型，并对所建模型预测效果进行验证和评价。方法收集2008年1月至2015年6月宁波市流感监测哨点医院 ILI 监测资料，对数据进行统计分析，建立ARIMA模型及ARIMA-GARCH模型对流感发病情况进行预测和评价。结果2008—2014年宁波市ILI累计报告101056例，发病率大致呈逐年下降趋势。针对ILI发病率的ARIMA模型构建中ARIMA（2，1，1）（1，1，1）12为最佳模型（BIC=6.250），白噪声残差分析得到Ljung-Box统计量Q值为6.027（P>0.05）。ARIMA-GARCH组合模型的预测效果较单一ARIMA模型理想，平均绝对误差分别为11.049和12.757。结论 ARIMA-GARCH模型可以模拟宁波地区流感的流行趋势，为流感防控策略的制定提供理论依据。

Forecast model analysis on the influenza-like illness in Ningbo, 2008-2015

Objective To establish forecasting model on the influenza-like illness (ILI) morbidity in Ningbo, and to examine and evaluate the applicability of the model. Methods Influenza surveillance data from January 2008 to June 2015 were obtained from ILI cases in Ningbo influenza monitoring hospitals. The ARIMA model and ARIMA-GARCH combined model were proposed to predict and evaluate the morbidity of influenza. Results During 2008 and 2014, 101 056 ILI cases were enrolled, and ILI morbidity rates had a decreasing trend year by year. ARIMA(2,1,1)(1,1,1)12 was chosen as the optimal model (BIC=6.250). The Ljung-Box statistical Q value was 6.027 by the white noise residual analysis(P>0.05). Comparative analyses showed that the ARIMA-GARCH combined modelwas more effective than the ARIMA model,and the mean absolute percentage error were 11.049 and 12.757, respectively. Conclusions ARIMA-GARCH model can be used to predict the epidemic trend of influenza successfully in Ningbo, providing a methodological basis for future influenza monitoring and control strategies.