气相色谱结合化学计量学分析4种食用植物油的指纹图谱

Analysis of Fingerprints of 4 Species of Edible Vegetable Oils by Chemometrics Combined with Gas Chromatography

运用气相色谱法对4类植物油(橄榄油、花生油、菜籽油和大豆油)的脂肪酸组成进行分析,并构建了植物油的指纹图谱,对4类植物油进行鉴别和分类。采用连续投影算法(SPA)对变量进行筛选,选出11个特征变量。以特征变量作为输入,使用主成分分析(PCA)和有监督模式识别(径向基函数神经网络(RBFANN)、线性判别分析(LDA)和最小二乘-支持向量机(LS-SVM))进行建模分析。结果表明,11个特征变量能够较好地区分4类植物油,PCA获得了较好的分类,RBF-ANN的预报结果最佳,预报率为92.6%,并且能准确预报二组分混合掺杂油样。该方法能够准确区分植物油种类,可用于食用植物油的鉴别和掺杂食用植物油的鉴定。

Gas chromatography（GC） was used to determinate the composition and contents of fatty acids in vegetable oils,including olive oil,peanut oil,rapeseed oil and soybean oil.And the GC fingerprint profile was employed for the fingerprint analysis and species classification of the four species of vegetable oils.11 feature variables were selected by successive projections algorithm（SPA）.Then,principal component analysis（PCA） and three supervised pattern recognition models： radial basis function artificial neural natwork（RBF-ANN）,least square-support vector machine（LS-SVM）,and linear discriminant analysis（LDA） were established to predict the species of the vegetable oils.The result demonstrated that the PCA obtained a clear clustering of objects respect to the species.RBF-ANN model performed better than the other two supervised pattern recognition models,with classification rate of 92.6%,and could predict the two component mixed oil sample accurately.The method could be used to distinguish the species of vegetable oil,and might be applicable for the identification of edible vegetable oils.