论文部分内容阅读
目的建立时间序列分析的自回归求和移动平均(ARIMA)模型,预测深圳市肾综合征出血热(HFRS)发病趋势。方法深圳市2005—2013年HFRS逐月发病率建立预测深圳市HFRS的最优ARIMA模型,利用2014年逐月HFRS发病率回代来检验模型预测效果,根据预测值与实际值的相对误差判断模型的预测精度,再以2005—2014年HFRS逐月发病率构建模型预测2015年的HFRS发病率。结果模型ARIMA(1,0,1)(1,0,1)12较好地拟合既往时间段的发病序列,各项参数(AR=0.993,MA=0.926,SAR=0.967,SMA=0.857)均有统计学意义(P<0.01),BIC值=-3.300,Ljung-Box模型统计量Q=20.794,P=0.107,模型残差为白噪声,2014年逐月HFRS发病率的预测值符合实际值的变动趋势,全年发病率预测值与实际值的相对误差率为20.74%。预测2015年深圳市HFRS发病率为4.28/100万。结论 ARIMA模型能很好地模拟深圳市HFRS发病率在时间序列上的变化趋势,并对未来的发病率进行预测。
Objective To establish an ARIMA model of time series analysis to predict the incidence of HFRS in Shenzhen. Methods The monthly incidence of HFRS from 2005 to 2013 in Shenzhen was used to establish the optimal ARIMA model for predicting HFRS in Shenzhen. The monthly incidence of HFRS in 2014 was used to test the model predictive effect. Based on the relative error between the predicted and actual values, The prediction accuracy of HFRS in 2015 was predicted from the 2005-2014 HFRS monthly incidence model. Results The model ARIMA (1, 0, 1) (1, 0, 1) 12 fitted the onset sequence and the parameters of the past time well (AR = 0.993, MA = 0.926, SAR = 0.967, SMA = 0.857) (P <0.01), the BIC value was -3.300, the statistic of Ljung-Box model was Q = 20.794, P = 0.107, and the model residual was white noise. The predicted incidence of monthly HFRS in 2014 was in line with the actual The relative error rate between the predicted value of the annual incidence and the actual value was 20.74%. The incidence of HFRS in Shenzhen is predicted to be 4.28 / 1 million in 2015. Conclusion The ARIMA model can well simulate the trend of HFRS morbidity in Shenzhen in time series and predict the future morbidity.