The concept of stochastic metamodeling of intelligent digital ecosystems

Abstract

The concept of stochastic metamodeling of intelligent digital ecosystems (IDE) is proposed – digital twins simulating organizational systems with a pronounced nature of knowledge transfer. The implemented methodology combines network analysis models (N approach) and probabilistic models of distributive semantics (S approach). The N approach includes percolation modeling (PM) and fractal analysis (FA) methods. The S approach includes probabilistic topic modeling (PTM) and stochastic trend forecasting model (SPM). All of the listed models can be solved by computational methods simulating parameters and variables by computer algorithms. During the sequential implementation of modeling algorithms, two tuples are formed, composed of sets of data parameters obtained during modeling – network Mn and semantic Ms. A formal meta-model of MM is presented, expressing the author's methodology of stochastic modeling, composed of the models Mn and Ms themselves, as well as hierarchical R_H, associative R_A and transitive R_T relations between the models. For an accurate solution of the descriptive problem of intellectual analysis of possible dependencies between variables in data sets, the concept of the DSM-method (plausible inference) of automated support of scientific research and methods of fuzzy set theory are proposed. The general results of testing the proposed approach to the study of the IDE based on processing data from citation networks of scientific publications and networks of scientific terms compiled from samples of bibliographic databases arXiv, Scopus and Google Scholar are considered.