Methods for Constructing Measurable and Metrizable Models of Knowledge Bases within Algebraic Approach

Abstract

Conventional methods to provide measurability and metrization of knowledge models cover only a small part of possible models, namely Bayesian networks and probabilistic analysis of models expressed by formulas of propositional calculus. The report proposes a new approach to building a wider class of measurable and metrizable knowledge models based on n-tuple algebra developed by the authors earlier. Besides. the proposed approach makes it possible to use clustering methods in models of knowledge bases.