三种预测模型在肺结核发病预测中的应用  

Application of Three Kinds of Model in Forecasting Prevalence of Pulmonary Tuberculosis

作　　者：易静[1] 胡代玉 杨德香 刘瑛 汪洋[1] 王润华[1] 

机构地区：[1]重庆医科大学,重庆市400016 [2]重庆市结核病防治所 [3]重庆市南岸区妇幼保健院

出　　处：《中国全科医学》2012年第13期1495-1497,共3页Chinese General Practice

基　　金：国家自然科学基金"基于人工神经网络的结核病疫情预测研究及软件实现"(30872160);重庆市科委自然科学基金计划资助项目(CSTC;2009BB5415)

摘　　要：目的建立合理的肺结核病发病预测模型,推测重庆市肺结核病疫情未来流行趋势,从而为合理分配卫生资源和持续有效地开展肺结核病防制工作提供科学依据。方法收集重庆市结核病防治所登记的1993—2008年肺结核年发病人数的登记资料,采用灰色预测模型、灰色马尔可夫组合预测模型与BP神经网络模型对重庆市结核发病人数进行预测对比分析,筛选最优拟合效果模型。结果采用灰色预测模型、灰色马尔可夫组合预测模型与BP神经网络模型对重庆市肺结核病发病疫情进行预测分析,3个模型的平均相对误差分别为23.81%、3.68%、3.52%。结论对于肺结核病发病的预测,BP神经网络模型拟合效果最好,预测精度更高,预测数据更合理。Objective To create a reasonable model to forecast the future epidemic trends of tuberculosis（TB） in Chongqing,so as to provides a scientific basis for rational distribution of health resources and sustainably effective work on pulmonary tuberculosis prevention and control. Methods The registration data were collected from Chongqing institute of Tuberculosis Prevention and Control about the yearly numbers of TB patients from 1993 to 2008.The grey system model [GM（1,1）],Grey-Markov Model,and Back Propagation Artificial Neural Network（BP-ANN） model were used to forecast the prevalence of tuberculosis in Chongqing.The predictabilities of these models were comparatively analyzed,and then the best fitting model would be chosen. Results The average relative errors of GM（1,1） model,Grey-Markov Model,and BP-ANN model were 23.81%,3.68%,and 3.52%,respectively. Conclusion BP-ANN model is the best fitting model to forecast the prevalence of tuberculosis.It is of a higher accuracy and more reasonable data of prediction.

关 键 词：肺结核 灰色预测模型 灰色马尔可夫 BP人工神经网络 

分 类 号：R521[医药卫生—内科学]