Chemometrics in qualitative analysis: identification, discrimination, and authentication

Non-targeted qualitative analysis has two aspects: instrumental, which involves collection of experimental data needed to solve a given problem, and mathematical, which involves analysis of these data to extract useful information and make reliable decisions. This review is devoted to the latter aspect, namely the general problems of chemometrics and machine learning, which are often incorrectly called artificial intelligence. Various problem formulations are considered, including discrimination and authentication; classification methods (binary, multi-class, or one-class); and types of made decisions (deterministic and probabilistic, soft and hard). Particular attention is given to analytical figures of merit such as sensitivity, specificity, and selectivity and how these characteristics are used for model optimization and validation. The concept of a cumulative analytical signal is also presented, and the prospects for its application in qualitative analysis are discussed. <br> The bibliography includes 162 references.