Research On Analysis And Optimization Method Of Feature Model Configuration Based On Stakeholders' Requests

Software product line(SPL) engineering is a paradigm to build families of similar software product in a specific way. The approach can significantly improve software development productivity, quality and time-to-market through sharing a common set of reusable resources often known as core assets. Feature models are a popular formalism for describing the commonality and variability of software product line in terms of features. Feature models symbolize a representation of the possible application configuration space, and can be customized based on specific domain requirements and stakeholders' goals. The challenges when using a feature model to derive a new SPL configuration are summarized as follows: 1) how to address the effective analysis and verification of software product line for large feature models that satisfy the properties of the behaviors, such as the request of system response; 2) how to find an optimized configuration based on extended feature model that minimizes an objective function, such as total cost.With respect to the effective analysis and verification of feature model configuration, this paper considered the software product line model checking theory and stakeholders' constraints. On the basis, this paper proposed a multi-valued software product line model checking method. Firstly, the method defined the slicing criterion which included the wanted features and excluded the non-relevant ones based on the stakeholders' requests, and computed the feature dependent set and independent set based on the semantic of the feature model. It then introduced three-valued logic to abstract the feature transition system based on the slicing result. Finally, based on multi-valued model checker, a case study was conducted to verify the abstract feature transition system and the results of the experiment illustrated the effectiveness of our approach.With regard to the optimized process of feature model configuration in SPLs based on the stakeholders' requests, this paper analyzed the existing optimization methods and constraints of the feature model. On the basis, this paper proposed an approach to optimize the feature selection based on the atomic set and improved genetic algorithm. Firstly, the approach simplified the feature model using atomic set and output an atomic set model. It then modeled the integrity constraints of the atomic set model which was employed as a fitness function to divide the valid configuration and invalid configuration. In the next step, it made use of uniform crossover operator for two parents which were selected from the valid and the invalid population separately. After crossover, mutation operation was used to generate new individuals from them. The evolutionary process continued until a chromosome satisfying the specific subject function was found. The fitness function of integrity constraints and the genetic operations together made the population to accelerate the convergence to the optimal solution.Finally, based on the presented feature model slicing method and the improved genetic algorithm, this paper designed and implemented an automated configuration tool which built a reliable configuration and optimized process based on stakeholder' requests. The tool supported cardinality-based feature modeling, feature model represented as trees, feature model slicing and configuration optimization based on feature diagrams.