Feature Location in Models (FLiM): Design Time and Runtime

In this chapter, we apply feature location to automate the identification and extraction of the features existing among a family of product models and re-engineering them into a model-based SPL. To address the feature location in software models (FLiM) challenge, we present two approaches: at design time (FLiMEA) and at runtime (FLiMRT). Both FLiMEA and FLiMRT approaches are different but complementary. FLiMEA takes information from design time models while FLiMRT takes information from runtime models. The FLiMEA approach combines Genetic Operations and Information Retrieval. Given a model and a description of a possible feature, model fragments extracted from the model are generated using genetic operation and are assessed using an information retrieval technique to rank the candidates based on the similarity with the feature description. The FLiMRT approach leverages the use of software architecture models at runtime. The information is collected in the software architecture model at runtime and each model element is assessed based on its similarity to the feature description. We evaluated our approaches in two real-world industrial case studies: BSH and CAF. The application of FLiMEA shows that the mean values of recall and precision are 72.99 per cent for BSH and 68.34 per cent for CAF while FLiMRT ranks the relevant elements in the top ten positions of the ranking in 84 per cent of the cases.