采用自回归移动平均模型预测我国狂犬病病例的研究

目的采用自回归移动平均模型（ARIMA）对我国大陆地区狂犬病月发病数进行预测，为我国狂犬病的防治工作提供参考依据。方法使用SPSS 19.0软件，利用2007年1月至2016年12月我国狂犬病的月发病数建立时间序列模型，并以2017年的月发病数为验证数据，评估和筛选最优模型，使用最优模型对2018年狂犬病流行趋势及发病数进行预测。结果最优模型为ARIMA（0,1,1）（2,1,0）₁₂，其平稳R2＝0.539，均方根误差＝17.653，Ljung-Box Q＝8.932，P＝0.881。 对2017年1—12月的数据进行预测，相对误差为1.55%，2017年我国狂犬病实际发病数为516例，预计2018年发病数将继续下降至398例。结论ARIMA（0,1,1）（2,1,0）₁₂模型能很好地拟合狂犬病发病的长期趋势和季节趋势，回代拟合和短期预测效果较理想。

Prediction of rabies cases in China by using autoregressive moving average model

ObjectiveTo predict the monthly incidence of rabies in the mainland of China by using autoregressive moving average model (ARIMA), and provide reference for the prevention and control of rabies in China.MethodsUsing SPSS 19.0 software, a time series model was established by using the monthly incidence data of rabies in China from January 2007 to December 2016, and the optimal model was validated by the monthly incidence data of rabies from January to December 2017. The optimal model was used to predict the incidence trend and case number of rabies in 2018.ResultsThe optimal model was ARIMA(0,1,1)(2,1,0)₁₂, with a stationary R2＝0.539, RMSE＝17.653, Ljung-Box Q＝8.932, P＝0.881. In predicting the data for January-December 2017, the relative error of prediction was 1.55%. A total of 516 rabies cases occurred actually in 2017. It was predicted that the case number of rabies in China would drop to 398 in 2018.ConclusionThe ARIMA(0,1,1)(2,1,0)₁₂ model can well fit the long-term trend and seasonal trend of rabies incidence, and the results of retrograde fitting and short-term prediction are ideal.