ARIMA模型在预测福建省新报告HIV/AIDS病例中的适用性  

Study on the adaptability of ARIMA model to forecast the trend of newly reported HIV/AIDS cases in Fujian province

作　　者：陈亮[1] 连巧龄 刘美增 张明雅[1] CHEN Liang;LIAN Qiaoling;LIU Meizeng;ZHANG Mingya(Department of Health Intervention in High-Risk Population,Fujian Provincial Center for Disease Control and Prevention,Fuzhou 350001,China)

机构地区：[1]福建省疾病预防控制中心高危人群干预工作研究室,福州350001

出　　处：《中国艾滋病性病》2021年第12期1353-1357,共5页Chinese Journal of Aids & STD

基　　金：福建省预防医学研究院建设科技创新平台建设项目(2019Y2001)。

摘　　要：目的利用福建省HIV/AIDS患者数据报告建立自回归移动平均模型(ARIMA),探讨数学模型在艾滋病疫情预测中的适用性。方法利用2010年1月1日至2020年12月31日福建省新报告的HIV/AIDS病例数据建立ARIMA模型,比较分析2020年HIV/AIDS病例报告情况并验证模型预测效果,运用拟合模型对2021年HIV/AIDS病例报告进行预测。结果 2010-2020年,福建省新报告HIV/AIDS病例呈现逐年缓慢上升趋势,经过数据序列平稳化处理和参数估计,确定ARIMA(1,1,2)(0,1,1)12为最优拟合模型。拟合优度检验显示贝叶斯信息准则(BIC)值为6.709,决定系数(R2)值为0.893,平均绝对误差百分比(MAPE)为10.709,Ljung-Box统计量为15.196(P>0.05)。残差序列符合正态分布特点,且残差的自相关系数和偏自相关系数都在95%CI范围内。模型预测曲线的变化趋势与实际趋势保持一致,预测2021年,福建省HIV/AIDS病例报告数为3 174例。结论 ARIMA模型可以应用于HIV/AIDS病例报告预测,可为防治政策制订和卫生资源配置提供重要参考依据。Objective To explore the adaptability of the mathematical model to forecast the trend of the AIDS epidemic by using monthly reported HIV/AIDS cases in Fujian province through the ARIMA model. Methods We established the ARIMA model based on newly reported HIV/AIDS cases in Fujian province during the period between Jan1, 2010, and Dec 31, 2020. We compared the actual number of HIV/AIDS cases reported in 2020 with the predicted value of the model and validated the effectiveness of the ARIMA model to predict the number of HIV/AIDS cases reported in2021. Results The number of newly reported HIV/AIDS cases was increasing steadily between 2010 and 2020. ARIMA(1,1,2)(0,1,1)12 model was identified as the optimal model after the sequential data was stabilized and parameter estimated.Test of goodness of fit indicated that the Bayesian Information Criterion, R2, MAPE, Ljung-Box Q value of the model was6.709, 0.893, 10.709 and 15.196(P>0.05), respectively. The residual sequence satisfied the normal distribution and the coefficient of Auto-Correlation and Partial Auto-Correlation lied in the 95% confidence intervals. The trend of the forecast curve was similar to the actual distribution pattern of the reported HIV/AIDS cases, the predicted number of the newly reported HIV/AIDS in Fujian province in 2021 would be 3174. Conclusion ARIMA model could be applied in the field of the forecast of HIV/AIDS case report and provide significant information for the policy development and resource allocation.

关 键 词：艾滋病病毒感染者/艾滋病患者 自回归移动平均模型 预测 

分 类 号：R512.91[医药卫生—内科学] R373.9[医药卫生—临床医学]