连续小波变换-支持向量回归-紫外分光光度法测定多组分B族维生素含量

Determination Multicomponent Vitamin B Using Continuous Wavelet Transform-Support Vector Regression-Ultraviolet Spectrometric Method

提出了连续小波变换-支持向量回归-紫外分光光度法(CWT-SVR-UV)多组分B族维生素含量测定方法.通过对维生素B混合样品测定光谱进行CWT处理,能够扣除光谱背景信号,降低光谱重叠与共线性的影响.将CWT-SVR-UV法用于维生素B1、维生素B2、维生素B3及维生素B6混合组分的同时测定,测定结果相对标准偏差分别为0.9%、1.8%、1.7%和0.5%.与基于原始光谱数据的SVR和偏最小二乘(PLS)建模方法相比较,本文提出的建模测定方法具有更好的预测准确度.

A new method for spectrometric determination of vitamin B using support vector regression （SVR） combination with continuous wavelet transform （CWT） was proposed. The ultraviolet （UV） spectral data of vitamin B and their mixtures were processed using CWT for removing of the spectral background. The calibration model was built using SVR based on the UV spectral data preproceased by CWT. It is shown that the accuracy of the prediction CWT-SVR-UV model was as good as or better than that of the models build based on the SVR-UV or that of the partial least square regression （PLS）-UV and CWT-PLS-UV. The CWT-SVR-UV model was used for determination of vitamin B1 （VB1）, vitamin B2 （VB2）, nicotinamide （VB3）, and vitamin B6 （VB6） with the relative standard error 0.9%, 1.8%, 1.7%, and 0.5%, respectively. CWT-SVR-UV is an effective method for determination of multicomponent vitamin B.