Machine learning in enhancing visualization of the spatial software architecture model

Visual aesthetics and computational efficiency are equally important aspects in the context of creating high-quality data representations for further analysis. A harmonious combination of these characteristics not only enhances the ease of perception but also optimizes data processing workflows, which is critically important in modern software systems. This paper proposes a novel combined approach for spatial graph placement. Special attention is given to spatial models of software architecture, which serve as key tools for visualizing complex relationships between components. The use of graph neural networks as a specialized heuristic forms the central element of this approach. Leveraging machine learning methods, the proposed solution improves visualization outcomes while enhancing computational efficiency. The application of a graph neural network ensures adaptability and enables the model to account for the specific features of the graph. In combination with a force-directed algorithm, this allows for maintaining a high level of visual aesthetics without significant increases in resource consumption. Thus, the new method offers a practical solution for effectively combining visual aesthetics with computational efficiency, representing an important step forward in enhancing the analysis of spatial models in software architecture.