Classification of pineapples from different origins by chemometric analyses on carotenoid profiles

Abstract

Geographical indication has played a vital role in sustaining the values of local products. From a chemical perspective, using the chemical profiles of products to differentiate them based on origins is highly useful due to its unbiased nature. Herein, we demonstrated that carotenoid profiles obtained from simple high-performance liquid chromatography (HPLC) analyses can effectively classify pineapples by their cultivars and origins. Principal component analysis (PCA), linear discriminant analysis (LDA), and self-organizing map (SOM) analysis were used to uncover interesting features of the data. Proper data pre-processing enabled PCA and LDA to differentiate pineapples by cultivars with prediction accuracies exceeding 90% in LDA. One-vs-all identifications of each pineapple origins achieved prediction accuracies above 70% with sufficient principal components (PCs). Importantly, SOM analysis significantly improved classification performance, increasing accuracies by at least 10%, and offered the added benefit of not requiring pre-selection of the numbers of PCs. Overall, this method has great potential for widespread adoption given the prevalence of carotenoids in various agricultural products.