2026-05-04T18:35:12.311Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:11855 2026-02-24T14:57:02Z 3

Applied Modified Metaheuristic: Machine Learning Tuned for Software Defect Detection, Chapter in CCIS Communications in Computer and Information Science: ASCIS 2025: Artificial Intelligence Based Smart and Secured Applications, Springer, volume 2820 2026 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/11855 https://link.springer.com/chapter/10.1007/978-3-032-17837-4_8 N. Bacanin V. Kostic R. Dragic L. Jovanovic M. Zivkovic B. Radomirovic V. Zeljkovic P. Spalevic D. Kavitha Software development involves structured methodologies encompassing coding, testing, debugging, and deployment, but faces challenges due to large codebases, diverse contributors, and high complexity. Manual testing and defect detection are increasingly inadequate, especially given the test oracle problem and coverage gaps. Artificial intelligence (AI) offers a promising alternative, enabling automated test case generation and defect identification. However, AI-based models struggle with imbalanced datasets, unstable software metrics, and the complexity of hyperparameter tuning—an NP-hard problem. To address this, metaheuristic algorithms, galvanized by nature, provide efficient search strategies for optimization. This study contributes by developing a modified metaheuristic algorithm for hyperparameter optimization, presenting a classifier for software defect detection, and proposing a supporting framework for continuous development and tuning. Real world testing using publicity available data suggest promising outcomes, with the top performing models tuned with the adapted algorithm attaining accuracy as high as .903846. bookPart Springer, Cham 120 130 10.1007/978-3-032-17837-4_8 CCIS Communications in Computer and Information Science: ASCIS 2025: Proceedings of International Conference on Advancements in Smart Computing and Information Security, volume 2820