Camellia oil grading adulteration detection using characteristic volatile components GC-MS fingerprints combined with chemometrics

Abstract

The volatile compounds combined with chemometrics, were used for different grade camellia oil adulterated detection. Using unsupervised models with fatty acids or triglycerides as input variables, the first-grade camellia oils (CAO Ⅰ) and second-grade camellia oils (CAO Ⅱ) occurred severely overlay in principle component analysis (PCA), for their almost quite identical compositions. While based on volatile components, hierarchical clustering analysis (HCA) presented a clear boundary between CAO Ⅰ and CAO Ⅱ samples. Subsequently, using our selected 15 volatile components by the variable importance in projection (VIP), both orthogonal projections to latent structure-discriminant analysis (OPLS-DA) and support vector machines (SVM), could be utilized as complementary models for camellia oil grade authentication with good classification rate (≥91.67%). In addition, according to those above characteristic variables with auto-scaling pretreatment, the optimized OPLS model could be recommended for adulterated level prediction (5%–100%, w/w).