Using Bayesian Belief Networks in Assessing Software Architectures

Quantitative techniques have traditionally been used to assess software architectures. We have found that early in the development process there is often insufficient quantitative information to perform such assessments. So far the only way to make assessments about an architecture, is to use qualitative assessment techniques like for instance a peer review. The problem with this type of assessment techniques is that they depend on the knowledge of the expert designers who use them. In this paper we introduce a technique that automates making qualitative assessments of software architectures.