Software architecture reconstruction: An approach based on combining graph clustering and partitioning

This article proposes an approach of improving the accuracy of automatic software architecture reconstruction. Many research uses clustering for the purpose of architectural reconstruction. Our work improves the results of coupling/cohesion driven clustering by combining it with a partitioning preprocessing that establishes a layering of the classes of the system. Two simple and not really efficient algorithms for software clustering are improved by applying this approach, as it is shown in the validation section.