Recent Papers

Software defect prediction (SDP) is a critical task in software engineering, aimed at improving software quality by identifying defective modules before deployment. Conventional defect detection techniques are frequently laborious and prone to human error. Automated defect prediction models have drawn a lot of attention as a result of improvements in Machine Learning (ML) classification algorithms. These models increase the effectiveness of defect management by using software metrics and historical defect data to categorise software modules as either defective or non-defective.In this study, we compare different classification models, such as Lazy-IBK, Lazy-K Star, SMO, Decision Stump, J48, and Nave Bayes, and measure the effect of ML classification algorithms on software defect prediction. Key performance metrics like accuracy, Kappa statistic, mean absolute error (MAE), root mean squared error (RMSE), relative absolute error (RAE), and root relative squared error (RRSE) are used in the study to assess these algorithms. The study develops and tests two hypotheses in order to validate the results: (H01) that ML classification algorithms have no discernible effect on defect prediction, and (H02) that there are no discernible differences between different classification models. The experimental findings show differences in error rates and predictive accuracy, emphasising the superiority of some models over others in the classification of software defects. Insights into the top-performing machine learning classifiers for defect prediction are provided by the research findings, which benefit software engineering by empowering developers to choose effective models for defect detection. To further increase prediction accuracy and dependability, future studies can investigate the combination of deep learning methods and hybrid models.

Copyright © 2025 Manuj Joshi. This is an open access article distributed under the Creative Commons Attribution License.