A framework for the comparison of nested software decompositions

Abstract

The evaluation of results obtained from software clustering algorithms has attracted the attention of many reverse engineering researchers. Several methods that compare flat decompositions of software systems have been presented in the literature. However, software clustering algorithms often produce nested decompositions. Converting nested decompositions to flat ones in order to compare them may remove significant information. We introduce a framework called END that reuses comparison methods for flat decompositions in order to compare nested decompositions without loss of information. We also present experimental results with END using several existing methods as plugins that demonstrate its usefulness.