One class classification (class modelling): State of the art and perspectives

Abstract

Classification, i.e., the prediction of one or more qualitative attributes of samples based on the measured data, is ubiquitous in chemistry, and, more specifically, in analytical chemistry. Among the possible classification strategies, class modelling techniques, which aim at describing one category at a time, present several advantages over discriminant ones, especially when dealing with asymmetric problems featuring one category of interest being well characterized and representatively sampled and another (made of everything that is not belonging to the first specific group) being under-represented by definition and highly heterogeneous.

In this review, the fundamentals of class modelling are illustrated, together with an overview of the main techniques of this kind proposed in the literature, namely Soft Independent Modelling of Class Analogy (SIMCA), Unequal Class Spaces (UNEQ), Potential Functions (PF), Partial Least Squares (PLS)-based algorithms, One Class-Support Vector Machines (OC-SVM) or Neural Networks (NN)-based strategies.