基于遗传算法结合最小二乘支持向量机的秦皮提取液快速定量分析方法

目的:探讨遗传算法(genetic algorithm,GA)结合最小二乘支持向量机(least squares support vector machine,LSSVM)用于秦皮提取液中多组分含量的快速测定的可行性。方法:以HPLC分析值作为参照,采集不同产地的秦皮提取液的紫外光谱,运用GA算法筛选光谱特征波长,再对筛选结果应用主成分分析(pricipal component analyze,PCA)进行特征提取,结合LSSVM算法建立快速测定秦皮甲素、秦皮乙素及秦皮素的数学模型。结果:应用GA算法筛选后,秦皮甲素、秦皮乙素、秦皮素的建模光谱维数分别下降为全谱的44%,44%,32%,所建模型的预测集相关系数(RP)分别为秦皮甲素0.948 6,秦皮乙素0.960 3,秦皮素0.929 3,预测均方根(RMSEP)分别为秦皮甲素0.085 2,秦皮乙素0.033 4,秦皮素0.012 1。结论:GA算法可以在充分保留光谱有效信息的基础上,显著降低模型的复杂度,结合LSSVM算法建立的模型可以用于快速测定秦皮提取液中秦皮甲素、秦皮乙素及秦皮素的含量,为中药提取液的质量控制提供新思路。

Fast determination of Cortex Fraxini extracts by GA-LSSVM

OBJECTIVE To explore the feasibility of using genetic algorithm(GA) combined with least squares support vector machine(LSSVM) for fast determination of Cortex Fraxini extracts.METHODS Ultraviolet spectra of Cortex Fraxini extracts from different batches were collected.The characteristic wavelength of spectral data was screened by GA and was compresssed by Pricinpal Component Analyze and subjected to LSSVM model to fast determine Aesculin,Aesculetin and Fraxetin of Cortex Fraxini extractions with the reference of HPLC.RESULTS Only 44% for Aesculin,44% for Aesculetin and 32% for Fraxetin wavelengths were selected by GA as significant wavelengths for determination.With the best calibration model built on wavelengths selected by GA,the correlation coefficient of prediction data(RP) for Aesculin,Aesculet and Fraxetin were 0.9486,0.9603,0.9293.And the root mean square errors of prediction data(RMSEP) for Aesculin,Aesculet and Fraxetin were 0.0852,0.0334 and 0.0121.CONCLUSION The number of wavelengths can be reduced by GA without significantly compromising prediction performance.It proves to be good in practice for fast determination of Aesculin,Aesculet and Fraxetin of Cortex Fraxini extractions by LSSVM.This method is advisable to control the quality of extractions of Cortex Fraxini.