摘要
采用虚拟组分自修正、自拟合的方法消除了干扰组分的影响,实现了人工神经网络(ANN)-紫外分光光度法不经分离的中药多组分浓度的同时测定。建立了包含训练网络和拟合网络的双网络ANN算法,提高了ANN算法的自学习、自判别能力,使复杂中草药体系多组分浓度预报的准确度大大提高。对21种不同来源的秦皮中秦皮甲素和秦皮乙素的含量进行了预测,预测结果与HPLC法相比较,以相对误差小于10%计,预测准确率大于90%。该法对秦皮甲素和秦皮乙素的测定精密度分别为0.37%,1.5%。
In the present paper, the simultaneous determination of multi-components in Chinese herbal medicine was performed by artificial neural network-UV spectrometry. The interference of other components was eliminated by self-revising and self-simulation of the virtual component. The double ANN including training and simulation network was established, and the capabilities of self-recognition and self-studying were improved. Therefore, the prediction accuracy of multicomponent content was improved greatly in the complicated Chinese herbal medicine system. The contents of aesculin and aesculetin, which were extracted from 21 Cortex Fraxinis, were predicted. Comparing the results with those of HPLC, the prediction accuracy was more than 90% within the relative errors less than 10%. The measurement preeisions of aeseulin and aeseuletin were 0. 37% and 1.5% respectively.
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2007年第1期126-130,共5页
Spectroscopy and Spectral Analysis
基金
国家自然科学基金项目(30371757)
吉林省科技厅项目(20030551-7)资助
关键词
人工神经网络
虚拟组分
秦皮甲素
秦皮乙素
Artificial neural network
Virtual components
Aesculin
Aesculetin
