COMPARATIVE ANALYSIS OF SOFTWARE REQUIREMENTS ENGINEERING METHODS

Abstract

Objective. To develop a clear and verifiable methodology for selecting methods for capturing software requirements and to determine the minimum sufficient set of artifacts for different tasks, with a separate focus on integration scenarios. Materials and Methods. The main artifacts of requirements engineering are considered as materials: SRS, user stories and acceptance criteria (including Gherkin), BPMN, UML (sequence and state diagrams), DFD/context diagram, ER model and data dictionary, as well as API contracts (OpenAPI, AsyncAPI). The comparison was made based on the criteria of clarity, verifiability, completeness/consistency, maintenance complexity, traceability to NFR, and suitability for integration, with a score on a scale of 0–5 and criterion weights; expert assessment was applied and the results were compiled into a “task/artifact” matrix and a final score table. Results. A matrix of task and artifact correspondence and a quantitative assessment confirming it were obtained. It was shown that for integrations, the most effective combination is “context + DFD + API contracts (OpenAPI/AsyncAPI) + UML sequence,” and when coordinating steps between systems, it is appropriate to add a BPMN fragment. BPMN is the leader for describing processes, UML state for the entity lifecycle, and the ER model and data dictionary for data and reporting. SRS demonstrates consistently high value due to its completeness and traceability to NFR, but does not replace behavioral and integration models. Conclusion. The proposed methodology helps to select artifacts for the task and avoid excessive documentation, adhering to the principle of a “minimally sufficient set.” This reduces the risk of inconsistencies in requirements and defects at system interfaces, increasing verifiability, traceability to NFR/SLA, and manageability of changes in development and integration projects.