全国人间狂犬病疫情的时间序列分析

摘要

In order to analyze the temporal distribution characteristics of human rabies in China and explore the incidence forecasting,R3.3.2 software were used to establish the optimal ARIMA model of monthly incidences from 2004 to 2015 in Chi-na.Then the monthly incidences from January to November 2016 were forecasted with the model and the prediction accuracy was evaluated.The annual incidences of human rabies in China were decreasing tendency since 2007,and the monthly incidence reached highly during August to October.The optimal model was ARIMA(2,1,1)(2,0,0)12and the mean absolute error (MASE)was 0.755.The mean relative error was 15.61% when predicting the incidences from January to November 2016.In conclusion,there exists seasonal variation for human rabies in China and ARIMA model can be applied for the short-term fore-casting.%目的 分析全国狂犬病疫情的时间分布特征,探讨用自回归移动平均模型(Autoregressive Integrated Moving Average Model,ARIMA)预测疫情发展趋势的可行性.方法 建立2004 ― 2015年全国狂犬病月度发病率时间序列,用 R3. 3.2软件建立最优 ARIMA模型,对2016年1~11月发病率进行预测,并评价预测效果.结果 2007年以来我国狂犬病年发病率呈下降趋势,8~10月为高峰季节;建立的最优模型为 ARIMA(2,1,1)(2,0,0)12,其平均绝对标准化误差(The mean ab-solute error,MASE)为0.755;2016年1~11月发病率预测结果显示,平均相对误差为15.61%.结论 我国狂犬病疫情存在季节性发病高峰,ARIMA模型能用于我国狂犬病疫情的短期预测.