Calculation of the Generalized Estimations in Sets of Features and their Interpretation

Abstract

A problem of searching the regularities (properties) on the basis of mapping results ofthe objects descriptions is studied on a subset of features on a numerical axis. A method isproposed for partitioning the features into sets by constructing graphs of connectednessof features. Generalized estimates by each set of features are interpreted as a latentquantitative feature for describing objects in the two-class recognition problem. Thedetected value characterizing regularity by the latent feature is regarded as a value of alinguistic variable and interpreted.