ARIMA模型在深圳市病毒性肝炎发病趋势预测的应用  

Application of ARIMA model in the prediction of viral hepatitis in Shenzhen city

作　　者：郑慧敏[1] 薛允莲[1] 黄燕飞[1] 戴传文[1] 姜世强[1] 

机构地区：[1]深圳市南山区疾病预防控制中心,广东深圳518054

出　　处：《中国热带医学》2015年第5期558-561,共4页China Tropical Medicine

基　　金：2014年度深圳市卫生计生系统科研项目(No.201402140)

摘　　要：目的通过探讨单纯求和自回归滑动平均模型（ARIMA模型）应用于病毒性肝炎发病率预测的可行性,为当前的防控工作提供科学依据。方法采用SAS9.2软件对深圳市2004~2013年的病毒性肝炎的月发病率进行ARIMA模型的建模拟合,预测2014年病毒性肝炎的月发病率,利用预测值和实际值的均方误差、平均绝对误差、平均绝对百分误差评价拟合效果,选择合适的模型预测深圳市2015年病毒性肝炎的月发病率。结果最终拟合为ARIMA（（12）,1,1）模型,残差为白噪声序列,预测值与实际值的均方误差为1.742,平均绝对误差为1.159,平均绝对百分误差为0.092,2015年深圳市病毒性肝炎发病率延续了自2011年以来的逐年上升的趋势。结论 ARIMA模型对病毒性肝炎的时间序列变动趋势的拟合效果较好,并对未来的发病率进行预测,可为病毒性肝炎防治提供科学依据。2015年预测结果提示病毒性肝炎的发病有上升的趋势,需要进一步调整相应防控策略。Objective To explore the feasibility of integrated autoregressive moving average（ARIMA）model in theprediction of the prevalence of viral hepatitis and to provide scientific basis for the control and prevention of viral hepatitis.Methods The ARIMA model was developed based on monthly incidence of viral hepatitis in Shenzhen during 2004-2013 with SAS 9.2 software and it was used to predict the monthly incidence of viral hepatitis in 2014.The mean square error,meanabsolute error and mean absolute percentage error of the predicted value and the actual value were measured to evaluatthemodel effect then a fitting model was chosen to predict the monthly incidence of viral hepatitis in 2015.Results The ARIMAmodel（（12）,1,1）was established finally and the residual sequence was a white noise sequence. Mean square error of thepredicted value and the actual value was 1.742, mean absolute error was 1.159 and mean absolute percentage error was 0.092.The incidence of viral hepatitis maintained the upward trend since 2011. Conclusions The change of time series of theprevalence of viral hepatitis can be simulated with ARIMA model, which can provide evidence for the prediction of viralhepatitis. The predicted values in 2015 suggested that the prevention and control strategies on viral hepatitis should beexplored further.

关 键 词：病毒性肝炎 时间序列 ARIMA模型 预测 

分 类 号：R512.6[医药卫生—内科学]