基于径向基函数神经网络的高血压分类诊断系统的建立  

Investigation on the Relationship between the Content of Some Trace Elements in Hair and Hypertension Disease Using RBF-NN

作　　者：李仲谨[1] 邱辉[1] 朱雷[1] 余丽丽[1] 张莎[1] 

机构地区：[1]陕西科技大学化学与化工学院,陕西西安710021

出　　处：《广东微量元素科学》2008年第12期14-19,共6页Trace Elements Science

摘　　要：为研究头发中Ca,Mg,A l,Cu,Zn 5种微量元素以及w(Zn)/w(Cu)与高血压的相关性,利用径向基神经网络(RBF-NN)的函数逼近、模式识别和分类能力强以及学习速度快等特点,对微量元素与高血压的相关性进行了研究;基于M atlab平台,对原始数据进行标准化预处理,45个作训练样本、8个作检测样本及其2个目标输出,建立了高血压分类的辅助诊断模型;同时与主成分分析法进行对照实验。结果表明,获得了最佳网络参数sc=0.1,m e=43,分类准确率达到96.226%,径向基神经网络在判别分类上优于主成分分析法。可见RBF-NN在揭示头发微量元素与高血压的相关性上是可行的,为临床高血压分类诊断提供了一种新的方法。To study the Correlation between the trace elements of Ca, Mg, Al, Cu, Zn in the hair and the ratio of Zn/Cu, the radial basis function neural network（ RBF- NN ）was employed to the correlation study between the trace elements and the hypertension because of the strong function approximation, the efficient pattern recognition, the accurate classification and the rapid learning speed of RBF. 53 original data were standardized, 45 data and 8 data were chosen as training samples and test sets, respectively, and then two target outputs were defined by radial basis function artificial neural network based on Matlab. A controlled experiment was designed as well. The results showed that the best net parameters （ sc = 0. 1 me = 43 ） were gotten, and the classification accuracy rate was 96. 226%. RBF is superior to principal component analysis in classification. It gains meaningful conclusions in a way of revealing the relationship between trace elements in hair and hypertension.

关 键 词：径向基函数神经网络 高血压 微量元素 

分 类 号：R544.1[医药卫生—心血管疾病]