A Generic Clustering Framework for Moose

Clustering helps with reengineering by gathering the software entities into meaningful and independent groups. The entities here can be any FAMIX entities, be it classes, methods, attributes etc. The affinity between two entities is calculated through the absolute difference of their properties and properties of the dependencies between the two entities; all the properties also have assigned weights. The clustering can be done by a range of clustering algorithms, including hierarchical and partitional algorithms. The result are groups of clusters, that can be examined through their quality metrics and if necessarily improved upon through another clustering run with adapted parameters. This paper describes generic clustering framework for the Object Oriented Reengineering Environmnet Moose, developed in the Software Composition Group at the University of Bern [1].