DESIGN SCIENCE RESEARCH APPROACHES TO BUILDING AN INTELLIGENT CONTRACT TESTING SYSTEM

Abstract

In the current climate of growing complexity in microservice information systems, developing effective approaches to ensuring their reliability and stability is particularly important.Objective. This article substantiates the application of the Design Science Research (DSR) methodology to the development of an intelligent system for testing microservice contracts. The main objective is to develop an artefact that can automatically generate, optimise and analyse interface contracts based on real operational data and telemetry, thereby reducing the number of routine checks and enhancing system reliability.Methodology. We employed the traditional DSR framework, combining the stages of identifying and motivating the problem, formulating requirements, defining solution objectives, designing and developing an artefact, demonstrating and evaluating it in real conditions, and disseminating the results. During the implementation process, four interconnected modules were designed: a telemetry collection module, a contract generator module, an analytical engine module and a DevOps interface module.Scientific novelty. For the first time, a unique combination of the DSR paradigm and intelligent adaptive testing mechanisms is proposed, offering automatic detection of contract changes with ≥ 95 % accuracy, prioritisation of tests using active learning while maintaining ≥ 95 % coverage of critical scenarios, and prediction of interaction 'hot spots' with a false alarm rate of ≤ 5 %. Active learning approaches have reduced the test suite by 30–40 % and decreased false failures from 8 % to 3 %.Conclusions. Experiments conducted on the bench with 15 microservices and 3,000 contract tests confirmed the system's effectiveness: CI pipeline time was reduced by 25 % and test volume by 30 %, while maintaining 98 % coverage of key contracts.Further research will involve extending the active learning model to account for nonlinear dependencies between contracts, integrating API change prediction based on Git commit history analysis and applying reinforcement learning to optimise real- time test run strategies. Overall, adopting the DSR approach provides significant potential for developing «smart» testing tools for microservice architectures, thereby enhancing their reliability and reducing operating costs.