Leonid V. Arshinskiy, Vadim S. Lebedev

Irkutsk State Transport University

It is shown that one of the ways to objectify productive knowledge bases in knowledge-based systems can serve as an inductive inference based on the combined method of similarity and difference and tables of joint occurrence of phenomena. An approach to the use of such a conclusion is proposed in conditions of possible low reliability and inconsistency of information sources forming tables. The approach is based on the concept of nonstrict probability, which, in turn, is based on the theory of logics with vector semantics in the VTF-logic variant. The tables themselves can be obtained from big data, primarily relational databases. It is assumed that such an approach will weaken subjectivism in the construction of production knowledge bases.

knowledge base, big data, inductive inference, nonstrict probability, logics with vector semantics