论文部分内容阅读
目的探讨自回归移动平均(ARIMA)模型在乙型肝炎(以下简称“乙肝”)发病数预测中的可行性,为预测乙肝发病趋势提供借鉴。方法利用“传染病报告信息管理系统”中湖北省2005―2016年乙肝分月发病数建立数据库,采用SPSS 12.0拟合ARIMA模型并进行预测。结果乙肝发病数原始数据存在一定的季节性及长期趋势,为非平稳时间序列;经差分处理为平稳序列后拟合得到ARIMA(0,1,1)(0,1,1)12为最优模型,残差检验为白噪声序列;模型回代性预测的平均绝对误差百分比为5.377%,<10%;前瞻性预测显示2016年各月份的实际值均在预测值95%置信区间内,且与实际值的平均相对误差仅为3.94%;模型预测2017和2018年乙肝年发病数分别为67 424、68 819例。结论 ARIMA模型能较好地模拟乙肝发病数在时间序列上的变动趋势,将其应用于乙肝发病趋势的预测是可行的。
Objective To investigate the feasibility of autoregressive moving average (ARIMA) model in predicting the incidence of hepatitis B (hereinafter referred to as “hepatitis B”) and provide references for predicting the incidence of hepatitis B. Methods The database of monthly incidence of hepatitis B in Hubei province during 2005-2016 was established by using the Infectious Diseases Reporting Information Management System, and the ARIMA model was fitted and predicted by using SPSS 12.0. Results There was a certain seasonal and long-term tendency in the incidence of hepatitis B virus, which was a non-stationary time series. After the differential treatment was a smooth sequence, ARIMA (0,1,1) (0,1,1) 12 was the best Model and the residuals were tested as white noise sequences. The average absolute error of model predictive regression was 5.377%, <10%. The prospective predictions showed that the actual values of each month in 2016 were all within the 95% confidence interval of the predicted values The average relative error from the actual value was only 3.94%. The model predicts the annual incidence of hepatitis B in 2017 and 2018 to be 67 424 and 68 819, respectively. Conclusion The ARIMA model can simulate the change trend of the number of hepatitis B in time series well, and it is feasible to apply it to the prediction of hepatitis B incidence trend.