An Adaptive Machine Learning Algorithm for Resilient Higher-Order Mutant Generation

In the field of software engineering, ensuring the reliability androbustness of software is paramount, and software testing plays a criticalrole in this process. Mutation testing, a fault-based technique, evaluates theeffectiveness of test suites by introducing artificial defects, known asmutants, into programs. This research presents a novel method forgenerating higher-order mutants (HOMs) using the Chemical ReactionOptimization (CRO) algorithm, which enhances the rigor of mutation testingby creating harder-to-detect mutants. The CRO algorithm employs fourtypes of collision operators: on-wall ineffective, synthesis, decomposition,and inter-molecular ineffective, to modify mutants and simulate complexfaults. Through experimentation with iterations set at 10, 30, and 50, it wasfound that increasing the number of iterations significantly reduces thenumber of mutants and increases their detection difficulty. Notably, with 50iterations, the approach achieved a 93% reduction in mutants and loweredthe mutation score to 27.77%, demonstrating the robustness of the generatedmutants. The research further introduces the HOMUsingCRO tool, whichautomates the mutant generation and testing process, generating XML-basedreports for effective mutant analysis. The proposed approach outperformsexisting techniques in both mutant reduction and mutation score, offering amore comprehensive solution for improving software test suiteeffectiveness.