Hybrid PSO-SA and DBSCAN with Mini-Batch K-Means for Efficient Software Test Clustering

Abstract

Software testing is the most crucial factor in making software reliable; old clustering methods alike K-Means and hierarchical clustering fail to deal with the complexities of large-scale test data. This is usually associated with poor scalability, ineffective test suite optimizations, and poor fault detection accuracy. Therefore, the present work strives to hybridize a cluster framework using PSO-SA with DBSCAN and Mini-Batch K-Means for efficient software test clustering to mitigate the limitations imposed by the aforementioned conventional clustering methods. Unlike conventional approaches, the proposed method utilizes density-based clustering for noise handling, batch-based refinement for scaling up, and swarm-intelligent optimized test case prioritization. Evaluation experiments done at the NASA Defect dataset demonstrated a fault detection rate of 95%, execution efficiency of 94%, and code coverage of 96%, higher than NNE-TCP as well as other baselines. The proposed method achieved 97 and 98 overall effectiveness and accuracy, respectively, validating the robustness of the approach to large scales of test clustering. The proposed method also improved effectiveness and valid into testing through comparison with NNE-TCP in standings of clustering accuracy, computational efficiency, and test prioritization. The framework helped enhance detection, reduced redundant test cases, and optimized test execution time. Integration into software testing environments will improve its automation overhead and reliability of software. Hence, it can be established as a valuable advancement in test suite clustering methodologies.