Supporting Decision Makers in Search-Based Product Line Architecture Design using Clustering

The Product Line Architecture (PLA) is one of the most important artifacts of a Software Product Line (SPL). PLA design can be formulated as an optimization problem with many factors. In this context, the MOA4PLA approach was proposed to optimize PLA design using search algorithms and metrics specific to the context. MOA4PLA treats the PLA design as a multi-objective optimization problem. At the end of the search process several PLA design alternatives are presented to the decision maker, difficulting the decision about which PLA alternative best fits with his/her needs. In this sense, this work proposes the usage of clustering algorithms to group PLAs design alternatives, according to their characteristics and assist the decision maker in the choice of one PLA design for the SPL. For this purpose, an empirical study was carried out, involving quantitative and qualitative experiments. Such an study integrated the K-Means++ and DBSCAN clustering algorithms in the MOA4PLA approach. The results of the experiments were promising, since an appropriate grouping of the solutions can be quantitatively observed, and also, qualitatively, the suitability of the solutions to the decision makers needs was verified.