An evolutionary methodology for optimized feature selection in software product lines

Feature modeling is the primary technology to capture and document the commonalities and variability among all of the members in a product line. Individual products are customized by selecting features according to the requirements. The work of feature selection is complex because of: 1) the complex dependencies and constraint relationship amongst features; 2) the multiple competing and conflicting non-functional requirements (NFRs); 3) the constraints to NFRs; 4) the explicit functional requirements. To select optimized feature set that conforms to the feature relations and satisfies both the functional and nonfunctional requirements and the related constraints, an evolutionary algorithm template which employs multi-objective optimization algorithms to optimally select features in SPLs, is proposed. In the experiments, two different algorithms are designed based on our template. Empirical results show the remarking performance of our algorithms on time especially when the feature models are large and complex.