自回归移动平均乘积季节模型在流行性脑脊髓膜炎发病预测中的应用

The Application of Autoregressive Integrated Moving Average Product Season Model on Prediction of Meningococcal Meningitis Cases

目的应用自回归移动平均(Autoregressive Integrated Moving Average,ARIMA)乘积季节模型(p,d,q)(P,D,Q)s,对流行性脑脊髓膜炎(流脑)发病数据的时间序列资料建模,并预测2010年流脑发病趋势,考察ARIMA乘积季节模型应用于流脑发病的预测效果。方法利用法定传染病报告系统的数据资料,采用Box-Jenkins方法建模,依据赤池信息量准则(Akaike's Information Criterion)和施瓦茨信息量准则(Schwarz's Information Criterion)结果确定模型阶数,用Box-Ljung统计量评价ARIMA模型的拟合效果,用所得模型对2010年流脑发病数进行预测。使用社会科学统计软件包时间序列分析模块对资料进行分析。结果对流脑的季节性时间序列建立了ARIMA(1,1,1)(0,1,1)12乘积模型,平均预测相对误差为3.09%,较好地拟合了流脑的发病趋势,并预测2010年全国流脑病例数为419例,在244～719例的可信范围内变动,发病高峰季节在3月份。结论 ARIMA乘积季节模型可较好地拟合流脑发病在时间序列上的变化趋势,是预测精度较高、效果较好的短期预测模型。

Objective To establish a model of Autoregressive Integrated Moving Average（ARIMA）product season（p,d,q）（P,D,Q）on time serial data of meningococcal meningitis cases,predict the possible cases in 2010,and evaluate the model predictive effect.Methods Using the data from National Notifiable Diseases Registry System（NNDRS）during 2000-2009,the method of Box-Jenkins was adopted to establish ARIMA product season model,the order of model was confirmed through Akaike Information Criterion（AIC）and Bayesian Information Criterion（BIC）,Statistics of Box-Ljung was used to evaluate the degree of fitness of ARIMA model,and data analyzed by Spss13.0.Results The model of product season ARIMA（1,1,1）（0,1,1）12 was established,average relative error of predict was 3.09%,model can appropriately fit the time series of meningococcal meningitis.Total 419 cases of meningococcal meningitis were predicted in 2010,Lower the 95% CI was 244-719,and the peak at March.Conclusion The product season ARIMA model can be used to fit the time series trend of meningococcal meningitis,and to predict the meningococcal meningitis cases with high prediction precision of short term time series.