Investigation of opportunities for test case selection optimisation based on similarity computation and search-based minimisation algorithms

Test Case Diversity investigations promise to reduce the number of Test Cases to be executed whereby addressing one of the drawbacks of automated model-based testing. Based on the assumption that more diverse Test Cases have a higher probability to fail, algorithms for distance analysis and search based minimisation techniques can help to enhance the quality of selection. This work discusses the application of Hamming Distance and Levenshtein Distance to compute similarity scores and outlines how Random Search and Hill Climbing can be applied to the problem of group optimisation based on pairwise Test Case similarity scores. The evaluation results, conducted with a test framework for automated test derivation and execution for IoT-based services, indicates that proposed Group Hill Climbing algorithm can outperform Random Search and at the same time utilising less computation time. The inclusion of the sequencebased Levenshtein algorithm shows advantages over the utilisation of the set-based Hamming-inspired scoring methodology.