A preliminary experimental study on optimal feature selection for product derivation using knapsack approximation

Software product lines (SPLs) technology produce software by integrating reusable software components based on customer requirements. Current researchers pay great attention to feature modeling technology that can represent SPLs' production requirements and functionalities. A key challenge is selecting valid and optimal feature combinations from the feature model to satisfy various requirements of customers and vendors, including various value and cost constraints. This paper experimentally studies a knapsack approximation algorithm of feature selection for automated product derivation in SPLs. Our approach generates an approximation solution by a modified Filtered Cartesian Flattening algorithm and obtains the optimal solution with a greed search. We performed experiments on randomly generated feature models with different characteristics. Experiments show that our approach can select highly optimal feature combinations effectively.