ARIMA模型在疟疾发病预测中的应用

Application of ARIMA model on prediction of malaria incidence

目的探讨ARIMA时间序列模型在疟疾发病预测中的应用,建立疟疾发病率的预测模型。方法基于1996～2005年安徽省怀远县的疟疾月发病资料,采用最大似然法估计模型参数,按照残差不相关原则、简洁原则确定模型结构,依据AIC与SBC准则确定模型的阶数,建立ARIMA疟疾预测模型。并用所得模型对2006年该县疟疾发病率进行预测。结果ARIMA(0,1,1)×(0,1,1)12模型能较好地拟合既往时间段上的发病率时间序列,其方差估计值为0.60,AIC=187.00,SBC=193.58,数学函数式为(1-B)(1-B12)Zt=(1-0.591B)(1-0.281B12)at。该模型对2006年月发病率的平均预测误差仅为0.03。结论ARIMA模型可较好地拟合疟疾发病在时间序列上的变化趋势,是一种精度较高的短期预测模型。

Objective To explore the applications of time series ARIMA model and fit the predictive model of malaria incidence. Methods ARIMA model was established based on the month malaria incidences from 1996 to 2005 in Huaiyuan County of Anhui Province. Parameters of the model were estimated through maximum likelihood method; the structure was determined according to criteria of residual un-correlation and concision, and the order of model was confirmed through Akaike Information Criterion （AIC） and Schwarz Bayesian Criterion （BSC）. The constructed model was used to predict the month incidence in 2006 and the result was compared with the actual incidence. Results ARIMA（0,1,1）×（0,1,1）12model can appropriately fit the malaria incidence with the least estimated variance of 0.60, AIC= 187.00, SBC= 193.58, and the mathematic formulation was （1-B）（1-B^12）Z, =（1-0. 591B）（1-0. 281B^12）at, The predicted month incidences were consistent with the actual values with the average estimated error of 0.03. Conclusion The ARIMA model can be used to fit the trend of malaria incidence in time series and to forecast for the malaria incidence with high prediction precision of short-term time series.