ABSTRACT
Classification methods are supervised; consequently, in order to predict whether an object belongs to a specific class, it is necessary to create a model based on the measurements collected on a training set of objects whose class-belonging is known in advance. Discriminant methods focus their attention on the dissimilarities among samples belonging to different classes, and operate by finding the decision boundaries that separate the regions of the multivariate space occupied by the categories under investigation. The theoretical description report suggests that both linear and quadratic discriminant approaches could be valuable tools for the analysis of chemical data. When chromatographic data are collected on a set of samples with the aim of obtaining a qualitative information, that is, with the scope of achieving a classification of the individuals, there are many tools available that differ among one another in complexity, applicability, and also in the kind of outcomes they produce.
