Date of Publication :8th October 2024

Abstract:Predicting software defects is crucial in software engineering since it helps to foresee problems, enhance software quality, and save development costs. This paper provides a comprehensive review of foundational and contemporary advancements in defect prediction methods. Early studies that focused on data mining static code properties were the first to construct predictive model s based on quantitative measures. Although benchmarking systems have facilitated the systematic evaluation of classification algorithms, systematic reviews have brought attention to the methodological transparency and rigor of fault prediction studies. Recent studies that concentrated on automated parameter optimization and the fusion of deep learning and ensemble approaches have greatly improved prediction accuracy. Comparative studies that have examined metric efficiency and cross-project assessments have brought attention to model transferability between software projects.

Reference :