Qualitative pattern recognition in chemistry: Theoretical background and practical guidelines

Qualitative pattern recognition methods find important applications in the chemometric sector to extract structured information from complex experimental data. Two main strategies can be distinguished: unsupervised analysis, aimed at investigating on the presence of groupings within the samples analysed, and supervised analysis, aimed at predicting the class membership of new samples. Supervised qualitative methods are, in turn, divided in two families: discriminant and class-modelling methods. The first ones require at least two classes to be defined, while the second ones are suitable also for one-class classification. The features of each strategy, with a focus on advantages and limitations, are described and compared. New trends in the methods, as well as recent attempts to force discriminant methods to behave as class-modelling ones, and vice versa, are also critically presented.

Automated summary

This study introduces qualitative pattern recognition methods in chemometrics, divided into unsupervised and supervised analysis. Supervised analysis is further divided into discriminant and class-modelling methods, with the former requiring at least two classes to be defined, and the latter suitable for one-class classification.