A methodology of automatic class diagrams generation from source code using Model-Driven Architecture and Machine Learning to achieve Energy Efficiency

Abstract

The automated generation of class diagrams is a crucial task in software engineering, facilitating the understanding, analysis, and documentation of complex software systems. Traditional manual approaches are time and energy consuming, error-prone, and lack consistency. To address these challenges, this research presents an automated proposed approach that utilizes Graph Neural Networks (GNNs), a machine learning algorithm, to generate class diagrams from source code within the context of Model Driven Architecture (MDA) and reverse engineering. A comprehensive case study is conducted to compare the results obtained from the automated approach with manually created class diagrams. The GNN model demonstrates high accuracy in capturing the system’s structure, associations, and relationships. Notably, the automated approach significantly reduces the time required for class diagram generation, leading to substantial time and energy savings. By advancing automated software documentation, this research contributes to more efficient software engineering practices. It promotes consistency, eliminates human errors, and enables software engineers to focus on higher-value tasks. Overall, the proposed approach showcases the potential of GNNs in automating class diagram generation and its practical benefits for software development and documentation.

Enhancing Class Diagram Dynamics: A Natural Language Approach with ChatGPT

“…Despite the prevalence of class diagrams in visualizing system structure and dynamics, their manual creation and maintenance remain time-consuming, error-prone, and resource-intensive [8][9][10]. This highlights an opportunity for AI-driven automation, specifically by dynamically enriching UML class diagrams based on detailed use case specifications expressed in natural language.…”