Answers are grounded in this paper and cite their sources.
By predicting which modules are likely to have defects before testing, software defect prediction helps keep development time and expenses down. Despite the prevalence of machine learning techniques, classification performance is frequently compromised in high-dimensional datasets by features that are not relevant to the problem. To solve this, heuristic algorithms are the way to go because effective feature selection is still NP-hard. This research presents a binary multi-objective starfish optimizer that finds the sweet spot between feature reduction and classification performance while selecting features. By combining many classifiers, an ensemble classifier based on Choquet fuzzy integrals can further improve the dependability of predictions. The method outperformed conventional classifiers in five separate NASA datasets that were used for validation. Defect prediction was discovered to be greatly affected by important software metrics as design complexity, number of branches, lines of code, and operators and operands count. Offering a trustworthy and interpretable approach for software defect prediction, the outcomes demonstrate that the suggested strategy enhances classification performance by 2% to 15% while preserving just 32% to 58% of feature set. For more general, high-dimensional classification jobs, our method shows great promise.
No takes yet. Share an insight, caveat, or question.
Chaitanya Kiran Mandadapu (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: