UDC 004.032.26
DOI: 10.36871/2618-9976.2020.12.001
Authors
Prokopchina Svetlana Vasil’evna
Doctor of Technical Sciences, Professor, Financial University under the Government of the Russian
Federation, Moscow, Russia
Abstract
The functioning of modern information systems takes place
in conditions of considerable uncertainty and heterogeneity
of information flows, which makes it difficult to process such
information: solutions are inaccurate, unstable, unfounded and
not interpreted, which has been repeatedly noted in the scientific
literature. This article suggests one of the methodological
ways to eliminate these difficulties based on the application
of the regularizing Bayesian approach (RBA) and Bayesian intelligent
measurements (BIM).
Information processing, both in the form of data and in the form
of knowledge, is performed in BIM systems on special scales,
scales with dynamic constraints that ensure the integration
of numerical and linguistic information based on a modified
Bayesian decision inference rule. Traceability of measurement
solutions is ensured by introducing a parallel branch of metrological
support for information processing. The semantic components
of conjugate scales of the SDC type implement the interpretability
(explainability) of measurements.
The introduction of metrological justification of knowledge
in technologies based on RBA provides a significant increase in
the efficiency of information processing, both in artificial intelligence
systems that traditionally work with knowledge, and
in measurement systems for the intellectualization of measurements
by ensuring the stability, interpretability and required
quality of the resulting solutions. In addition, it is the metrological
support of information technologies that will allow for the reliable
integration of artificial intelligence and measurement technologies,
which is an urgent need of modern information systems.
Keywords
Uncertainty
Intelligent measurement
Knowledge metrology