Enhancing Harmony Search Metaheuristic Algorithm for Coverage Efficiency, Test Suite Reduction, and Running Time in Combinatorial Interaction Testing

Optimization has developed powerful algorithms for solving complex problems efficiently. The effectiveness of these algorithms, also called metaheuristics, largely depends on the capabilities of their search techniques. Combinatorial Interaction Testing (CIT) is an efficient technique for detecting faults caused by interactions among system factors. By systematically covering combinations of input values, CIT can uncover faults that arise from complex interactions. One key aspect of CIT is coverage efficiency, which ensures that the desired interaction coverage is achieved with the fewest possible test cases, optimizing the testing process. Additionally, test suite reduction and execution time plays a crucial role by minimizing the number of test cases and time respectively without significantly compromising critical coverage, thereby improving the overall efficiency of the testing process. The Harmony Search (HS) algorithm is a recent metaheuristic that mimics musicians’ improvisation to create pleasant compositions according to aesthetic standards widely used in different areas continuously adjusts solution variables to find the best outcome. However, several studies show that most metaheuristics including HS face challenges when adjusting parameters to improve their performance. In this paper, we introduce an improved version of Harmony Search, referred to as eHS, and utilize CIT techniques for coverage efficiency, test suite reduction, and execution time. A significant shortcoming of the algorithm is improper arrangement of the harmony memory and inadequate exploitation during improvisation when trying to solve complex problems. In order to improve algorithm efficiency, we specifically embedded one-parameter-at-a-time approach of generating CIT test case within the harmony memory and adjust the improvisation processes to accommodate CIT test suite reduction, execution time, and coverage efficiency performance respectively. Despite HS’s tendency to get stuck in local optima, we dynamically adjust its parameter values in our proposed eHS. The experimental results demonstrate that eHS outperforms the other algorithms for CIT in terms of coverage efficiency and test suite reduction. Specifically, eHS achieved the best CIT interaction coverage efficiency performance with 15 (60.00%) successes out of 25 entries, 30 out of 33 cases (90.91%) for CIT test suite size performance, and unfortunately, 00.00% for CIT execution time performance. This study concludes that eHS emerges as a valuable algorithm for generating CIT test suite. Finally, we proposed a potential possibilities for further exploration of HS for CIT.