Multivariate Classification Techniques

In the present article, supervised qualitative data modeling is presented as a fundamental branch of pattern recognition, including two main approaches: discriminant classification and class-modeling. The choice of the correct strategy is influenced by the final aim: the discriminant approach is appropriate when at least two classes are meaningfully defined, while class-modeling is recommendable when a single class is on the focus. Key aspects related to the development, optimization and validation of suitable qualitative models are presented, offering to the reader proper guidelines to face this branch of chemometrics.