ABSTRACT
Detection of linear sub-clusters is the topic addressed in this chapter. A fuzzy class may contain linear sub-clusters, which may be detected using a prototype-based partitional method. The Generalized Fuzzy n-Lines (GFNL) algorithm is such an alternating optimization procedure. Other prototypes considered are linear varieties. Clusters having a degree of linearity are also considered. A different clustering method is based on principal components of a fuzzy class. Principal components are defined and their use for clustering purposes is considered.
