Метод автоматизованої ідентифікації обмежених контекстів під час проєктування систем електронної комерції

Abstract

The article examines the issue of automating the strategic stage of e-commerce system design, focusing on modeling complex business logic and structuring the domain. Modern e-commerce systems are characterized by a high level of complexity in business rules and a large number of interconnected processes. Effective management of this complexity is possible through the application of the Domain-Driven Design approach, which involves decomposing the system into logically separated modules known as Bounded Contexts. The quality of identifying these contexts directly affects the system’s viability, the clarity of the codebase, and its potential for further evolution. Traditional approaches to identifying context boundaries are based mainly on heuristic methods and expert sessions, which are resource-intensive and dependent on the human factor. Existing formal methods that rely on structural data analysis are often unable to correctly interpret the semantic nuances of business terminology [1]. This paper proposes an information technology approach that uses the capabilities of generative artificial intelligence and large language models (LLMs) for automated analysis of the problem space. The main focus of the article is the development of a method for semantic clustering of requirements, which makes it possible to identify hidden linguistic patterns in the description of business processes. The proposed approach involves using LLMs to analyze the project’s Ubiquitous Language and to form a Context Map based on the semantic similarity of concepts rather than only their technical relationships. An algorithm is described that transforms textual specifications and user stories into formalized domain models, determining the recommended boundaries of responsibility for each module. The research results demonstrate that applying a generative approach makes it possible to significantly increase the objectivity of domain modeling, minimize the cognitive load on architects, and ensure validation of the system’s logical integrity at the early stages of design. The developed technology serves as an intelligent decision-support tool, enabling the creation of flexible and business-change-adaptive e-commerce models.