Using Genetic Algorithms to Modularize Conceptual Models

Conceptual models often evolve into large and monolithic artifacts. This is a threat to comprehensibility and maintainability as the size of the resulting artefacts exceeds the cognitive processing capabilities of human beings. Breaking down monoliths into a modular structure is an established technique in software and systems engineering.

Task

Part of this project is to realize Genetic Algorithms (GA) to automatically derive modularizations of given monolithic models. From a conceptual perspective, this project involves the challenge of encoding the problem and solution domain into a format, that can be used as an input for the GA. Moreover, a fitness function – most likely a multi-objective one – needs to be defined, based on domain-specific metrics, that yields the GA toward the good solutions. From a technological perspective, this project can extend an existing EMF and Jenetics based application or build a new one.

Possible applications can be on Entity Relationship models or other languages like BPMN, DMN, and ArchiMate.

Further Reading (Excerpt)

• Bork, Dominik, Antonio Garmendia, and Manuel Wimmer. “Towards a Multi-Objective Modularization Approach for Entity-Relationship Models.” ER Forum, Demo and Posters 2020. CEUR-WS. org, 2020. http://ceur-ws.org/Vol-2716/paper4.pdf https://model-engineering.info/img/ModulER-Presentation.pdf
• https://jenetics.io/
• https://github.com/jku-win-se/module-eer
• Moody, D. L., & Flitman, A. (1999). A methodology for clustering entity relationship models - a human information processing approach. In International Conference on Conceptual Modeling (pp. 114-130). Springer, Berlin, Heidelberg.