Due to different users' requirements, contemporary software has become feature-rich in terms of input functions (i.e., parameters) and selections (i.e., values). Exhaustive testing on sophisticated software systems is impractical as far as testing time and cost are concerned. Various software testing techniques such as equivalence class partitioning, boundary value analysis and decision table, have been proposed in the literature with the sole purpose of detecting single-factor faults. Unlike earlier works, combinatorial t-way testing (where t indicates the interaction/combination strength) supports the detection of faults caused by two or more input system parameter interactions (i.e., multi-factor faults) while efficiently minimizing the size of the test suite as compared to exhaustive test lists. Over the past few years, metaheuristic optimization algorithms have appeared to be the most common choice for researchers since their effectiveness proves to offer optimal/near-optimal results. However, as the t-way test suite generation is a Non Deterministic Polynomial Time Hard (NP-hard problem), no single t-way strategy can guarantee superiority to others for all types of system configurations. Hence, this thesis presents a new uniform strength (fixed interaction strength) t-way strategy based on Gravitational Search Algorithm (GSA), called Gravitational Search Test Suite Generator (GSTSG). The key contribution of this research thesis is that the proposed GSTSG strategy employs GSA for the first time for t-way test data generation, which has yet to be explored in software testing research. The benchmarking results showcase that GSTSG obtains competitive results in most system configurations when compared to other existing strategies and addresses higher interaction coverages of up to t = 10 owing to the memoryless feature of GSA. Overall, GSTSG produces optimal test suite sizes in 56.25% of system configurations out of a total of 64 benchmarked configurations. Moreover, the results obtained by GSTSG is statistically significant from other strategies in all multiple comparisons using the Friedman test and also shows differences in 9 paired tests among 15 total pairwise comparisons according to the Wilcoxon Signed Ranks test.
