Applicability analysis of generative adversarial neural networks in the tasks of system design modeling

This thesis explores the applicability of Generative Neural Networks in the tasks of system design modeling. The research addresses the key role of the design phase in the software development lifecycle, emphasizing the importance of modeling for visualizing and formalizing system requirements. Despite the widespread use of graphical models like UML diagrams to describe structural and behavioral aspects of systems, modeling often receives insufficient attention due to its complexity and the significant time investment required. Generative neural networks offer a promising solution to these issues by leveraging deep learning algorithms capable of understanding and generating data. This potential for automation can significantly reduce the effort and time needed for system design, a traditionally human-intensive task. However, the application of neural networks in system design remains an underexplored area, necessitating further investigation. The primary objective of this study is to evaluate the effectiveness of using generative models for automating the system modeling process. The study concludes that while generative AI shows substantial promise for automating UML diagram generation, there are limitations and challenges that need to be addressed. These include the necessity for further refinement of generative models and the importance of integrating domain-specific knowledge to enhance the quality of automated designs. The research contributes to the ongoing development of automated design tools, providing a foundation for future advancements in the field. ​​