邱琳, 郁会莲, 李红蕾, 朱妮, 贠鹏飞. 求和自回归移动平均模型在陕西省细菌性痢疾发病预测中的应用[J]. 疾病监测, 2014, 29(5): 403-406. DOI: 10.3784/j.issn.1003-9961.2014.05.017
引用本文: 邱琳, 郁会莲, 李红蕾, 朱妮, 贠鹏飞. 求和自回归移动平均模型在陕西省细菌性痢疾发病预测中的应用[J]. 疾病监测, 2014, 29(5): 403-406. DOI: 10.3784/j.issn.1003-9961.2014.05.017
QIU Lin, YU Hui-lian, LI Hong-lei, ZHU Ni, YUN Peng-fei. Application of autoregressive integrated moving average model in predicting incidence of bacillary dysentery in Shaanxi[J]. Disease Surveillance, 2014, 29(5): 403-406. DOI: 10.3784/j.issn.1003-9961.2014.05.017
Citation: QIU Lin, YU Hui-lian, LI Hong-lei, ZHU Ni, YUN Peng-fei. Application of autoregressive integrated moving average model in predicting incidence of bacillary dysentery in Shaanxi[J]. Disease Surveillance, 2014, 29(5): 403-406. DOI: 10.3784/j.issn.1003-9961.2014.05.017

求和自回归移动平均模型在陕西省细菌性痢疾发病预测中的应用

Application of autoregressive integrated moving average model in predicting incidence of bacillary dysentery in Shaanxi

  • 摘要: 目的 探讨时间序列模型预测传染性疾病发病率的可行性,应用自回归移动平均(autoregressive integrated moving average,ARIMA)模型对陕西省细菌性痢疾进行预测,为制定细菌性痢疾防治策略提供依据。 方法 根据2004-2012年陕西省细菌性痢疾月报告发病率的时间序列,以2013年1-12月的月发病率作为验证数据,建立ARIMA模型,并对预测效果进行评价。 结果 陕西省2004-2012年细菌性痢疾月发病率即含有长期递减趋势又含有以年为周期的季节效应,拟合的相对最佳模型为ARIMA(0,1,1)(1,1,0)12。残差分析统计量经检验差异无统计学意义(Ljung-Box Q=21.994,P=0.143),提示残差为白噪声。2013年1-12月实际值与预测值的相对误差平均值为20.75%,最大40.37%,最小4.94%。 结论 ARIMA模型可以较好地预测陕西省细菌性痢疾的发病趋势,模型预测效果的优化有待原始数据的进一步积累。

     

    Abstract: Objective To evaluate the feasibility of time series model to predict the incidence of infectious diseases. Methods According to the time series of reported monthly incidence of bacillary dysentery in Shaanxi province from 2004 to 2012, the autoregressive integrated moving average (ARIMA) model was established by using the incidence data of bacillary dysentery from January to December 2013 as demonstration data. The predictive power of ARIMA model was evaluated. Results The case curve is not only with a long-term descending trend but also with annual seasonality. The relative optimum fitting model was ARIMA(0,1,1)(1,1,0)12. Ljung-Box Q had no statistical significance (Ljung-Box Q=21.994,P=0.143) and residuals was the white noise. The average of the relative error between actual value and predicted value from January to December in 2013 was 20.75% (maximum 40.37%, minimum 4.94%). Conclusion The ARIMA model can be used to effectively predict the incidence of bacillary dysentery in Shaanxi. More original data are needed in order to optimize the model.

     

/

返回文章
返回