Tool Support for Learning Architectural Guidance Models from Architectural Design Decision Models

Abstract

This paper presents an approach to architectural knowledge management that does not assume existing architectural design decisions or pattern applications are documented as architectural knowledge, but benefits from more existing data. We drew inspiration from manual qualitative research methods for mining patterns and architectural knowledge and created a guideline model of which the ADD models are instances. We evaluated our approach on 11 cases from the gray literature. We found that it can provide suitable recommendations after modeling only a single case and reaches theoretical saturation and recommendations with low to very low errors after only 6-8 cases. Our approach shows that creating a reusable architectural design space is possible based only on limited case data. Our approach not only provides a novel approach to architectural knowledge management but can also be used as a tool for pattern mining.