中药材三七提取液近红外光谱的支持向量机回归校正方法  

A Method for Analysis of Extracting Solution of Panax Notoginseng Using Near-infrared Spectroscopy and Support Vector Regression

作　　者：瞿海斌[1] 刘晓宣[1] 程翼宇[1] 

机构地区：[1]浙江大学药物信息学研究所,杭州310027

出　　处：《高等学校化学学报》2004年第1期39-43,共5页Chemical Journal of Chinese Universities

基　　金：国家"十五"重大科技攻关项目 (批准号 :2 0 0 1BA70 1A0 1);国家重点基础研究发展规划项目 (批准号 :G19990 54405;G1999054408)资助

摘　　要：提出近红外光谱的支持向量机回归校正建模方法 .以中药材三七渗漉提取液为实际分析对象 ,对其近红外光谱数据进行预处理和主成分分析后 ,用支持向量机回归算法建立人参皂苷 Rg1,Rb1,Rd以及三七总皂苷的近红外光谱校正模型 .以 Rg1,Rb1和 Rd的 HPLC测定值及三七总皂苷的比色法测定值为参照 ,将本文方法与偏最小二乘回归和径向基神经网络建模方法相比较 ,结果表明 ,本文所建模型的预测准确性优于后两者 ,可推广应用于中药提取过程的近红外光谱分析 .A new method for building near-infrared spectroscopy(NIRS) calibration model using support vector regression(SVR) was proposed. After NIRS data of the extracting solution of Panax notoginseng herb were processed with pretreatment and principal component analysis, NIRS calibration models of ginsenoside R g1, R b1, R d and panax notoginseng saponins(PNS) were built by using SVR. The reference measurements of ginsenoside were performed by HPLC method for ginsenoside R g1, R b1, R d and colourimetric method for PNS. The proposed method was compared with partial least square regression(PLSR) and radial-basis function neural network(RBFNN) modeling methods. The results showed that the predictive accuracy of NIR calibration models built by SVR was much better than that of the models built by PLSR and RBFNN. Therefore, the method could be applied to NIR analysis for extraction process of Chinese medicine.

关 键 词：中药材 三七 提取液 近红外光谱 支持向量机 回归分析 校正方法 过程分析化学 

分 类 号：R284.2[医药卫生—中药学]