ARIMA模型在河南省AIDS疫情预测中的应用*
Application of ARIMA models in forecasting epidemic trend of AIDS in Henan Province

目的：探讨ARIMA模型在河南省AIDS疫情预测中的可行性。方法：收集河南省2005年至2013年AIDS发病率和病死率数据，分别建立ARIMA模型，通过模型拟合优度统计量筛选最优模型，采用平均误差率评估模型预测效果，并用所构建的模型预测2014年至2016年河南省AIDS疫情。结果：河南省AIDS发病率和病死率的最优ARIMA模型分别为ARIMA(0,1,1)和ARIMA(0,2,1)；模型的拟合值均符合其实际流行趋势，平均误差率分别为0.16和0.25；预测2014年至2016年河南省AIDS发病率分别为3.40/10万、3.91/10万和4.50/10万，病死率分别为15.34%、8.82%和7.64%。结论：ARIMA模型很好地模拟了河南省AIDS疫情的演变趋势，可用于河南省AIDS发病率和病死率的预测。
Aim: To explore the feasibility of autoregressive integrated moving average(ARIMA) model in forecasting epidemic trend of AIDS in Henan Province. Methods: ARIMA model was trained with the annual morbidity and fatality rate of AIDS from 2005 to 2013 in Henan Province, respectively. Moreover, the goodness？？of？？fit was used to determine relative optimal model, and the mean error rate(MER) was calculated to assess the accuracy of prediction models. Finally, the data of 2014 to 2016 were forecasted by the best？？fitting ARIMA models.Results: ARIMA (0, 1, 1) model was the best model for morbidity of AIDS, while ARIMA (0, 2, 1) was the optimal models for fatality rate of AIDS. The fitting trend of prediction model was consistent with the observed data, and the MER of two optimum ARIMA models was 0.16 and 0.25 respectively. The forecasted value of the morbidity from 2014 to 2016 in Henan Province was 3.40, 3.91 and 4.50 per 100 000 populations and the predicated value of fatality rate was 15.34%, 8.82% and 7.64%. Conclusion: The ARIMA model fit epidemic trend of AIDS well and can be used in predication of AIDS morbidity and fatality rate of Henan Province