Facilitating requirements inspection with search-based selection of diverse use case scenarios

Use case scenarios are often used for conducting requirements inspection and other relevant downstream activities. While working with industrial partners, we discovered that an automated solution is required for optimally selecting a subset of use case scenarios, aiming to enable cost-effective requirements inspection. In this paper, relying on a natural language based use case modeling methodology to specify requirements as use case models and derive use case scenarios automatically, we propose a search based and similarity function based approach to optimally select most diverse use case scenarios from the ones automatically generated from the use case models. We conducted an empirical study to evaluate the performance of various search algorithms together with eight similarity functions, through an industrial case study and six case studies from the literature. Results show that the search algorithms significantly outperformed Random Search and (1+1) Evolutionary Algorithm together with the Normalized Longest Common Subsequence (NLCS) similarity function performed significantly better than the other 31 combinations of the search algorithms and similarity functions for most of the problems.