Chemometrics in Laboratory Data Analysis

Food is a complex system that requires holistic quality assessment. Chemometrics uses mathematical, statistical, and computer science methods to analyze and interpret chemical data, which means good prospects for food quality evaluation. This review covered Russian and international publications indexed in Scopus, PubMed, MEDLINE, Web of Knowledge, Google Scholar, IEEE Xplore, Science Direct, and eLIBRARY.RU (RSCI). The search queries included such keywords as chemometrics; chemometric methods; principal component analysis; PLS (projection to latent structures); artificial neural network (ANN); multivariate classification; multivariate data analysis. The main chemometric tools applied to food systems included hierarchical cluster analysis (HCA), principal component analysis (PCA), latent structures-discriminant analysis (PLS-DA), projections to latent structures (PLS), quadratic projection to latent structures (QPLS), multiple linear regression (MLR), artificial neural network (ANN), support vector machine (SVM), k-nearest neighbors (KNN), and ensemble model prediction (RF, XGBoost). The PCA proved to be the most popular chemometric method applied in the food industry. However, combinations of methods were always more effective than a single one. The KNN methods appeared to be quite unreliable. Combinations of chemometric methods demonstrate the best prospects, e.g., PCA + PLS-DA + ANN or PCA + PLS-DA +KNN. If combined with instrumental tools, they are able to improve analytical accuracy and provide effective management approaches, thus ensuring sustainable food industry.