Method for analyzing the behavior of mathematical models of the intelligent systems design process under conditions of multilevel uncertainty

Abstract

The article proposes a method for analyzing the behavior of mathematical models of the design process of intelligent systems under conditions of multilevel uncertainty. The method is intended for quantitative assessment of changes in the characteristics of the design process under the influence of epistemic, structural and semantic components of uncertainty. To describe the state of the mathematical model, a system of partial criteria was used that characterize the completeness of the requirements reflection, the correctness of the grammatical-mathematical transformation, structural consistency, adaptability and resilience of the design process. Partial criteria are combined with an integral quality indicator and an integral risk indicator, which allows for a comprehensive assessment of the behavior of the mathematical model under changes in design conditions. The relationship between the components of multilevel uncertainty and the criteria for evaluating the design process is established, which allows determining the nature of their direct and indirect impact. An experimental study was conducted for two intelligent systems with different manifestations of uncertainty. Based on the variation of structural coherence and adaptability, the change in the integral quality of the design process was investigated and the areas of stable behavior of the mathematical model were determined. The results obtained confirm the possibility of using the proposed method for a comprehensive assessment of the behavior of mathematical models of the design process and determining the need for their adaptation under changing uncertainty conditions.