ARIMA模型在新疆流感样病例占门急诊病例百分比预测中的应用

ARIMA model in prediction of influenza-like illness rate in Xinjiang

目的探讨应用时间序列自回归滑动平均求和模型（ARIMA）在建立流感样病例占门急诊病例百分比（ILI%）预测模型方面的应用。方法利用新疆2012—2014年每周17家哨点医院的ILI占门急诊病例百分比（ILI%）数据建立ARIMA模型,拟合ILI%的变化趋势,用残差序列分析进行模型诊断,并对2015年1~26周（上半年）ILI%进行预测,来评价ARIMA模型的预测效果。结果 2012—2014年新疆的ILI%是周期性变化,经模型诊断发现ARIMA（1,0,1）（0,1,0）52模型为最优模型,通过对2015年1~26周数据的外推,预测值与实际值的平均相对误差为8.75%,且实际值均在预测值的95%可信区间内,实际值与预测值变动趋势一致。结论 ARIMA模型可对ILI%进行很好的拟合,可作为新疆维吾尔自治区ILI%短期预测模型。

[Objective]To explore the application of Autoregressive Integrated Moving Average（ARIMA）model in prediction of influenza-like illness rate（ILI%）.[Methods]ARIMA model was established based on the data of weekly influenza like illness rates in 17 sentinel hospitals of Xinjiang from 2012-2014,to fit the change trend of ILI%,and the model diagnosis was performed using residual analysis. The predictive capacity of the ARIMA model was tested by forecasting the ILI% in first half of 2015（1st-26 th week）.[Results]There was periodic variation in ILI% in Xinjiang from 2012-2014,and ARIMA（1,0,1）（0,1,0）52was chosen as the optimal model.The relative prediction error of forecast values to actual values was 8.75%. The actual values were in the 95% confidence interval of the forecast values,and the actual value had analogical change tendency with the predicted value.[Conclusion]ARIMA is a good tool for modeling and forecasting the ILI%,which can be used as the short-term forecasting model of ILI% in Xinjiang Uygur Autonomous Region.