Applying enhanced graph clustering to software dependency analysis

Dependencies between classes give key information about the static structure of an object oriented software system. For industrially sized systems it is difficult for the developer to visually analyze the dependencies between classes and to detect patterns of dependencies that frequently occur throughout UML class diagrams. In this paper, automatically detecting dependency patterns in software designs is focused. After applying graph clustering techniques to dependency graphs extracted from class diagrams it has been found that these techniques were not able to detect key dependency patterns An algorithm is proposed to detect such dependencieswhich also improves on the studied graph clustering techniques when applied to dependency analysis of class diagrams.