目的 探讨乘法季节回归求和移动平均模型(ARIMA)在安徽省细菌性痢疾预测中应用,为细菌性痢疾的预测提供参考。方法 根据2010～2015年每月细菌性痢疾发病数据,运用R软件拟合乘法季节性ARIMA模型,并对2016年细菌性痢疾的发病数进行预测。结果 安徽省细菌性痢疾预测最优模型为ARIMA(0,1,0)×(0,1,1)₁₂模型,残差统计量检验差异无统计学意义(Ljung-Box =12.812,P=0.383),提示残差为白噪声,模型拟合值与实际值平均绝对误差百分比为10.30%, 2015年1～12月预测值和实际值平均绝对误差百分比为11.89%。结论 建立的乘法季节ARIMA模型能很好地拟合安徽省细菌性痢疾的变动趋势,模型预测效果较好,可用于安徽省细菌性痢疾发病短期预测。

Objective To explore the application of multiple seasonal autoregressive integrated moving average model in predicting bacillary dysentery incidence in Anhui province,and provide evidence for the forecast and early warning of bacillary dysentery. Methods According to the 2010~2015 year data of monthly bacillary dysentery in Anhui province,R software was used to fit the multiplicative seasonal ARIMA model,and the number of bacillary dysentery in 2016 was predicted. Results Multiple seasonal ARIMA(0,1,0)×(0,1,1)₁₂was established for the prediction of bacillary dysentery incidence in Anhui province. There was no significance in Ljung-Box Q(Ljung-Box Q=12.812,P=0.383)and residuals was the white noise. The average absolute percentage error between fitted and value actual was 10.30%. The average absolute error percentage of the predicted and actual values from January to December 2015 was 11.89%. Conclusion The ARIMA model fitting effects are well and suitable to short-term prediction of bacillary dysentery incidence in Anhui province.