New trends in qualitative analysis: Performance, optimization, and validation of multi-class and soft models

Abstract This article considers current trends and challenges in qualitative analysis. The focus is not on classification methods per se, but rather on their general features, such as the way the classification results are presented (hard or soft classification), or how the method is applied (single-class or multi-class). We put forward the main problems, which have not yet been fully investigated and, therefore, require special attention. Such problems include definitions of the main figures of merit in cases of multi-class, one-class, and soft and hard classifications; optimization methods in cases of rigorous and compliant approaches to classification; well-known validation methods such as (test set and cross-validation), as well as a recently introduced Procrustes cross-validation. At each point, we focus on novel ideas and methods, some of which have not yet been fully investigated and tested in practice. We believe that they can create new trends in qualitative analysis and, therefore, deserve to be discussed.