Abstract

Discovering linear dependencies in data sets is discussed in the paper as a part of data mining approach. The proposed method is based on the minimization of a special type of convex and piecewise linear (CPL) criterion functions defined on a given data set C. The division of the set C into a family of linearly dependent clusters Ck allows to form a family of local regression type models. As a result, each subset Ck can be characterized Ck by its own linear model. The K-plans algorithm which is similar to the K-means algorithm can be used for dividing the set C into a family of linearly dependent clusters Ck Also a different approach to this problem, based on the CPL criterion functions is discussed here.