Application of multivariate machine learning methods to investigate organic compound content of different pepper spices

Abstract

The aim of this study was to uncover all aspects and extract comprehensive and valuable information from the data obtained from different pepper varieties using machine learning (ML) methods. The red pepper (RP), fabricated isot (FI), and customary isot (CI) spices were stored for 12 months and the variations in the organic compound content were monitored every 3 months. The data set has been subjected to a supervised ML method Random Forest (RF), unsupervised ML methods principal component analysis (PCA), t-Distributed stochastic neighbor embedding (t-SNE), and hierarchical cluster analysis (HCA). The classification accuracy yielded by the RF model was 100%. RF model showed that terpenoids, acids, and alkanes were ineffective in identifying the differences between pepper spices, but glucose, succinic acid, citric acid, and fructose were primarily responsible for the variations between pepper spices. FI peppers differed significantly from other pepper spices in terms of their chemical compositions. Although most organic compounds exhibited positive correlations; furan-fructose, furan-glucose, furan-citric acid, and glucose-malic acid showed negative correlations. RP peppers were mostly stable for the first 6 months of storage, but after this month, due to changes in malic acid, aldehyde, glucose, and fructose, they displayed similar properties as CI. The organic compound content of CI peppers rapidly changed in the first 3 months of storage and stayed almost stable for the remaining 9 months. Various ML methods were effectively employed in this study to examine the changes that different pepper spices exhibited in association with storage. Practical applicationsPeppers, which are often used in food products for the purpose of bitterness, flavor, and color are mostly consumed after drying. The dried pepper is stored for a long time before consuming. During storage, there are significant changes in the organic compound composition and pepper quality. In this study, revealing the changes in organic compound content in detail with machine learning methods can aid researchers and the pepper industry as a decision-making tool.