时间序列分析——广义回归神经网络组合模型在痢疾发病率预测中的应用探讨  

Discussion on Application of ARIMA-GRNN Combination Model in Shigella Epidemic Situation Forecast

作　　者：杨小兵[1] 徐勇[1] 彭磊[1] 贺圆圆[1] 魏巍[1] 刘军[1] 严薇荣[2] 

机构地区：[1]宜昌市疾病预防控制中心,湖北宜昌443000 [2]华中科技大学同济医学院流行病与卫生统计学系,湖北武汉430030

出　　处：《中国初级卫生保健》2008年第8期64-66,共3页Chinese Primary Health Care

摘　　要：目的探讨组合模型在痢疾流行预测及防治效果评价中的应用价值。方法根据2001--2006年宜昌市痢疾月发病率建立痢疾发病率组合预测模型，并对2007年痢疾月发病情况进行预测。结果宜昌市痢疾发病率组合预测模型平均相对误差为O．078，较单一时间序列分析模型降低了74.34％；对2007年痢疾月发病率进行外推预测显示．5-10月份发病率较高。结论组合模型拟合效果较为理想，是一种短期内预测精度较高的预测模型，在疾病预防控制领域中具有重要的应用价值。OBJECTIVE To explore the application value of combination model in Shigella epidemic situation forecast and the effect of prevention and control. METHODS According to the monthly Shigella incidence of Yichang City from 200l to 2006, a forecasting mod- el of monthly Shigella incidence was established, and the trend of the monthly incidence of Shigella in 2007 was forecasted in order to evaluate the effect of prevention and treatment. RESULTS The average relative error of combination model for Shigella is 0.078, which is less than the error of single ARIMA model. The results show that the incidence rates between May and October are high. CONCLU- SION Combination model with a good fitness and high prediction precision in short-term time series is valuable in the field disease control and prevention.

关 键 词：组合模型 痢疾 预测 

分 类 号：R51[医药卫生—内科学]