Architecture of multi-agent systems for generative automatic matching among heterogeneous systems

Abstract

This paper presents the generative automatic matching (GAM) approach, implemented through a multi-agent system (MAS), to address the challenges of heterogeneity across meta-models. GAM integrates automatic meta-model matching with model generation, offering a comprehensive solution to complex systems involving diverse architectures. The key innovation lies in its ability to automate both the detection of correspondences and the transformation of models, improving the precision and recall of matching processes. The system's scalability and adaptability are enhanced by MAS, allowing for efficient management of diverse meta-models. The approach was evaluated through relational to big data UML meta-models (RBDU) case study. The results demonstrated high accuracy, with precision and recall metrics approaching 1, underscoring the robustness of GAM in managing heterogeneous systems. Compared to traditional methods, GAM offers significant advantages, including automated matching and generation, adaptability to various domains, and superior performance metrics. The study contributes to the field of model-driven engineering (MDE) by formalizing a method that effectively bridges the gap between heterogeneous meta-models. Future research will focus on refining matching heuristics, expanding case studies.