Personalized recommender systems for software product line configurations
Juliana Arriel · Digitalen Hochschulbibliothek Sachsen-Anhalt (Universitäts- und Landesbibliothek Sachsen-Anhalt) · 2018
Software Product Lines (SPLs) have been employed in the industry as a mass customization process that reduces production costs and time-to-market. However, the inherent complexity and variability of SPLs lead to an exponentially growing amount of possible products. Thus, especially when dealing with large SPLs, scalability and performance concerns start to be an issue and specialized assistance becomes crucial to guide decision makers during product configuration. In this context, SPL configuration has been a hot research topic in the last years. In this thesis, we provide an overview on SPL configuration techniques and, based on our insights, employ recommendation techniques to enable an efficient SPL configuration process and provide accurate and scalable solutions to decision makers. To this end, we offer four main contributions. First, we adapt state-of-the-art collaborative-based recommender algorithms to the SPL configuration context. Second, we consider non-functional properties to improve the efficiency and quality of the recommendations. Third, we propose an advanced recommender system that relies also on contextual information to enable reconfiguration at runtime. Fourth, we provide visual support to guide decision makers through an easy and comprehensive configuration process by allowing them to focus on a limited set of valid and relevant parts of the configuration space. We empirically demonstrate the usability of the implemented algorithms and tool in different real-world scenarios from different domains.