Segmentation and classification. An application to patients’ risk estimation

In many fields of research we have to deal with the so called “classification problem” namely that of assigning an individual x to one of several pre-specified classes ω j , j = 1 … r. That assignment is based on some features or measurements made on the individual x, Q i (x), i = 1 … p. This problem is difficult because there often exists a substantial amount of variability in the measurements of individuals belonging to the same class.