A hybrid approach to suggest software product line portfolios

Graphical abstractDisplay Omitted HighlightsThe work proposes a hybrid approach to deal with the Product Portfolio Scope Problem.The approach is composed by a solution to deploy the feature relevance indicated by the customers into code assets of a SPL, based on a systematic method (SQFD).The approach includes a method to estimate the cost of an asset based on common and relevant measures related to source code, together with a fuzzy system to deal with the imprecision to set reference values.The work presents an application of an NSGA-II to search for products minimizing the cost and maximizing the relevance of the candidate products.The approach was evaluated using different scenarios, exploring the mains aspects related to method in the practice: size, granularity of features and products search space.The previous version of our hybrid approach was dependent on the employed technologies and algorithms. Herein we reformulate our approach, detaching it from any particular technique/algorithm.The data collection process associated with our approach was improved to facilitate the hybrid approach's usage and mitigate associated construct validity threats.A more comprehensive evaluation, which focused on show the real word usefulness and scalability of our hybrid approach. To validate the usefulness of our approach, it was used the SPL associated with a tool broadly employed in both industrial and academic contexts (ArgoUML-SPL). The scalability of our approach was evaluated using a synthetic SPL.All the experiments were based on the guidelines defined by Arcuri and Briand in order to evaluate the statistical significance of this kind of work. Software product line (SPL) development is a new approach to software engineering which aims at the development of a whole range of products. However, as long as SPL can be useful, there are many challenges regarding the use of that approach. One of the main problems which hinders the adoption of software product line (SPL) is the complexity regarding product management. In that context, we can remark the scoping problem. One of the existent ways to deal with scoping is the product portfolio scoping (PPS). PPS aims to define the products that should be developed as well as their key features. In general, that approach is driven by marketing aspects, like cost of the product and customer satisfaction. Defining a product portfolio by using the many different available aspects is a NP-hard problem. This work presents an improved hybrid approach to solve the feature model selection problem, aiming at supporting product portfolio scoping. The proposal is based in a hybrid approach not dependent on any particular algorithm/technology. We have evaluated the usefulness and scalability of our approach using one real SPL (ArgoUML-SPL) and synthetic SPLs. As per the evaluation results, our approach is both useful from a practitioner's perspective and scalable.