ARIMA季节乘积模型在肠道传染病预测中的应用  

Model of multiple seasonal ARIMA and its application on prediction of enteric infectious disease

作　　者：张翼飞[1] 陈洪 刘岭[1] 张彦琦[1] 郭波涛[1] 易东[1] 

机构地区：[1]第三军医大学统计学教研室,重庆400038 [2]重庆光学机械研究所,重庆401123

出　　处：《激光杂志》2008年第2期91-91,共1页Laser Journal

摘　　要：目的:建立细菌性痢疾月发病数的预测模型,探讨AR1MA季节乘积模型在时间序列资料分析中的应用。方法:采用非条件最小二乘法估计模型参数,通过季节差分方法使原始序列平稳,按照残差不相关原则、简洁原则确定模型结构,依据AIC和SBC准则确定模型阶数,建立ARIMA预测模型。结果:对所分析的季节性时间序列建立了乘积ARIMA(0,0,1)(0,1,1)12模型。方差估计值为288.106,AIC=619.661,SBC=620.492。对模型进行白噪声残差分析(p=0.632),拟合优度统计量表明ARIMA的估计具体模型为:(1-B12)Zt=(1-0.34B)(1-0.559B12)αt是适合的。结论:通过ARIMA(0,0,1)(0,1,1)12模型与ARIMA(0,1,1)12模型对细菌性痢疾月发病数预测效果的比较,表明ARIMA季节乘积模型是一种短期预测精度较高的预测模型。Objective：To establish a model of multiple seasonal autoregressive integrated moving average （ARIMA） （p, d, q） （P, D, Q） s on month - morbility of Bacillary Dysentery , and to explore the applications of multiple seasonal ARIMA model. Methods： The parameters of model were got based on unconditional least squares. The primitive series may become steady by logarithmic transformation and finite difference. The structure is determined according to criteria of residual un - correlation and concision. The order of model was confirmed through Akaike Information Criterion and Schwarz Bayesian Criterion. So ARIMA predictive model was fitted. Results： For the data of Bacillary Dysentery, the model of ARIMA （ 0,0, 1 ） （ 0,1,1 ） 12 was established. In this model, the estimation of variance is 288.106, AIC = 619. 661, SBC = 620. 492. The white- noise residual was analyzed based on the residual analysis（ p = 0. 632 ）. Aceording to the rich table, it shows that the best ARIMA model is（ 1 - B^12）Z^t = （1 -0.34B）（ 1 -0. 559B^12 ）α1. Conclusion： The model of ARIMA can be used to forecast incidence of Bacillary Dysentery. And it has a high prediction precision for short - term time series.

关 键 词：ARIMA季节乘积模型 时间序列 肠道传染病 细菌性痢疾 

分 类 号：O211.61[理学—概率论与数理统计] O211.67[理学—数学]