基于BP人工神经网络的病毒性肝炎发病率自回归预测模型  

BP Artificial Neural Network Based Predicative Model for Viral Hepatitis'Incidence

作　　者：关鹏[1] 何苗[2] 周宝森[1] 

机构地区：[1]中国医科大学流行病学教研室,110001 [2]中国医科大学附属第一医院计算机室

出　　处：《中国卫生统计》2003年第5期269-271,共3页Chinese Journal of Health Statistics

基　　金：国家自然科学基金资助 ( 39770 6 75和 30 1 70 833)

摘　　要：目的　研究BP人工神经网络在存在自回归现象的病毒性肝炎发病率预测中的应用。方法　首先通过时间序列分析 ,发现病毒性肝炎发病率存在一阶自回归现象 AR(1) ,然后以AR(1)和平均气压、平均气温、平均降水量和平均蒸发量四个气象指标为网络输入 ,将 1979～ 1994年某地病毒性肝炎发病率变换值作为网络输出 ,选择 1979～ 1992年数据 ,利用MATLAB软件 6 5的神经网络工具箱 4 0构建、训练与模拟网络 ,预测另外两年病毒性肝炎发病率 ,建立病毒性肝炎自回归BP人工神经网络模型。结果　BP人工神经网络非线性相关系数RNL=0 80 ,而线性自回归模型RNL=0 62。结论　BP人工神经网络较好地描述了自回归模型中的非线性关系 ,内插及外推预测均较好 ,可以用于解决存在自回归现象的问题 ,效果好于传统方法。Objective To study the application of BP artificial neural network tools for the forecasting of viral hepatitis incidence which exists autoregression phenomenon.Methods First, time series analysis was adopted, The network input variables included AR(1), average air pressure, average air temperature, average precipitation and amount of evaporation. The output of network is a transform value of the incidence rate of viral hepatitis Neural network tools box 4.0 of Software MATLAB 6 5 was used to structure, train and simulate BP Artificial Neural Network. The data from 1979 to 1992 were used as a training set and the data from the remaining 2 years made up the test set.Results The application of the BP artificial neural network enabled the R\-\{NL\} of 0 80 , while the R\-\{NL\} of linear model is 0 62.Conclusion BP artificial neural network is superior to conventional methods in solving problem which exist Autoregression phenomenon.

关 键 词：BP人工神经网络 病毒性肝炎 发病率 自回归 预测模型 流行病学 

分 类 号：R181.3[医药卫生—流行病学]