A quantitative, evidence-based approach for recommending software modules
Thaís Burity, Glêdson Elias · 2015
In distributed software product line projects, dependencies between components influence on communication and coordination needs among their respective development teams. As an alternative to reduce such needs, it seems interesting to cluster tightly-coupled components into loosely-coupled modules as long as each module is developed by a single team. In such a context, considering that numerous clustering possibilities exist, this paper presents a quantitative, evidence-based approach for recommending software modules by clustering software components. The proposed approach has a threefold foundation: quantitative measures that characterize dependencies between components; a search-based clustering algorithm to recommend modules; and a quantitative measure that characterize dependencies between modules, which can guide the allocation of development teams to modules.