Towards the characterization of realistic models

Abstract

Custom generators of graph-based models are used in MDE for many purposes such as functional testing and performance benchmarking of modeling environments to ensure the correctness and scalability of tools. However, while existing generators may generate large models in increasing size, these models are claimed to be simple and synthetic, which hinders their credibility for industrial and research benchmarking purposes. But how to characterize a realistic model used in software and systems engineering? This question is investigated in the paper by collecting over 17 different widely used graph metrics taken from other disciplines (e.g. network theory) and evaluating them on 83 instance models originating from six modeling domains. Our preliminary results show that certain metrics are similar within a domain, but differ greatly between domains, which makes them suitable input for future instance model generators to derive more realistic models.

Digital Tools in UMLS Metathesaurus Knowledge Processing

“…The primary objective in multigraph processing is to range concepts by significance degree in searching task context. G. Szarnyas and coauthors have evaluated some graph metrics before [5]. In the current research, we have adapted one of the most powerful metrics known as clustering coefficient.…”

Change Impact Analysis Case Study for Aviation: Mutation Testing

“…In the study where a generative model of software dependency graphs was designed (GD-GNC) and the structure of artificial graphs created with this model were compared with real software graphs, it was examined that there is a common arrangement in software dependency graphs [12]. In the study conducted to characterize the graph-based model to be used in software and systems engineering, evaluations were made on many graph metrics and sample models [18]. Automatic generation of graph models has become an important approach in various applications of software and systems engineering.…”