论文部分内容阅读
目的探索建立适合于流感样病例预测的自回归求和移动平均模型(ARIMA模型)。方法采集深圳市南山区2006—2011年流感样病例监测数据,绘制序列图,差分使序列平稳化,通过自相关分析和偏相关分析进行模型识别,根据AIC(赤池信息准则)和BIC(贝叶斯信息准则)确定模型参数,建立ARIMA预测模型,用Q统计量法对模型适用性进行检验,用2012年全年实际监测数据与模型预测值进行比较,评价模型预测效果。结果 2006—2011年流感样病例累计报告199 360例,月发病最大值9 765例,月发病最小值594例,平均月发病2 769例。通过对2006—2011年各月的监测数据进行分析发现,各年度流感样病例发病呈现明显的高峰和低谷,高峰在每年5—8月份,低谷在当年的11月份至次年2月份,不同年度略有波动。对序列进行一阶差分后可得到较为平稳的序列,适合进行模型拟合,经过模型拟合诊断发现ARIMA(0,1,1)×(0,0,1)12模型为最优模型,AIC值和BIC值最小,分别为1 239.19和1 245.98,Box-Ljung检验结果 Q值为19.07,P>0.05,通过2012年拟合值与实际值比较,结果差异无统计学意义(P>0.05)。结论 ARIMA模型可以较好地对流感样病例进行拟合分析预测。
Objective To establish an autoregressive and moving average model (ARIMA model) suitable for prediction of influenza-like illness. Methods The surveillance data of influenza cases from 2006 to 2011 in Nanshan District of Shenzhen City were collected, and sequence diagrams were drawn. The sequences were smoothed by the difference, and the models were identified by autocorrelation analysis and partial correlation analysis. According to the AIC (Chi Chi Information Criteria) and BIC Sri Lanka’s information criteria) to determine the model parameters, the establishment of ARIMA prediction model, the use of Q statistics to test the applicability of the model, with the 2012 annual actual monitoring data and model predictive values were compared to evaluate the model predictive effect. Results A total of 199 360 flu-like cases were reported from 2006 to 2011, with a monthly incidence of 9 765 cases, a monthly minimum of 594 cases and an average monthly incidence of 2 769 cases. Through the monitoring data of each month from 2006 to 2011, we found that the incidence of influenza-like cases in each year showed obvious peaks and valleys, the peak is in May-August each year, the trough is in November of the current year to February of the following year, in different years Slight fluctuations. After the first-order difference of the sequence, a more stable sequence can be obtained, which is fit for model fitting. After the model fitting diagnosis, ARIMA (0,1,1) × (0,0,1) 12 model is the best model, AIC The values of Box-Ljung test were 19.07 and P> 0.05, respectively. The results of Box-Ljung test showed that there was no significant difference between the fitted value and the actual value in 2012 (P> 0.05) . Conclusion ARIMA model can be a good flu-like cases were fitted to predict the prediction.