基于动态学习比率BP神经网络的时间序列预测方法  

Predictive Method of Nonlinear Time Series Based on Dynamic Error Correction Using BP Neural Network

作　　者：丁守銮[1] 王洁贞[1] 胡平[1] 

机构地区：[1]山东大学卫生统计教研室,250012

出　　处：《中国卫生统计》2002年第4期194-198,共5页Chinese Journal of Health Statistics

基　　金：山东省自然科学基金资助课题 (编号Y2 0 0 0C1 9)

摘　　要：目的　探讨人工神经网络在时间序列资料分析中的应用。方法　利用动态学习比率BP算法以双曲正切函数为功能函数的非线性时间序列预测方法。结果　建立HFRS发病率的两种ANN预测模型 ,其预测精度远远高于传统方法。结论　BP人工神经网络可以用于疾病发病率或死亡率的预测。Objective To explore predictive method of nonlinear time series based on using BP neural network.Methods Based on dynamic error correction BP using artificial neural network with Hyperbolic Tangent function as activation function and the number of hidden node six.Results Build two kinds of ANN forecast models of HFRS incidence They are better than traditional method on prediction precision.Conclusion BP artificial neural network can be used to forecast for disease incidence or mortality.

关 键 词：BP人工神经网络 非线性时间序列 动态学习比率 发病率 医学统计学 

分 类 号：R195[医药卫生—卫生统计学]