SOFTWARE DEFECT PREDICTION MODELS FOR MITIGATING THE RISK OF BUG RECURRENCE IN LINES OF CODES: A SYSTEMATIC REVIEW
Keywords:
Feature selection techniques, Software defect prediction, Software metrics, Systematic literature reviewAbstract
The custom of recent research in the field of software defect prediction is to display frameworks that provide software quality assurance teams with the ability to concentrate purely on software defect-prone codes. This has enabled software development time and maintenance activities to be managed more efficiently. Previous studies involved intricate software defect prediction datasets and methods. However, more detailed and inclusive insight into the state-of-the-art software defect prediction research method is needed. This systematic literature review aimed at evaluating the trends in the research field, the algorithms deployed, the methods utilised, the datasets used, the feature selection techniques mostly applied, the software metrics used, and the evaluation measures used in software defect prediction studies between 2017 and June 2023. Upon the application of selection and rejection criteria, 104 studies published on software defect prediction from January 2017 to June 2023 were evaluated. It was shown that 82 published articles on software defect prediction applied classification techniques, or about 78% of the total study. Meanwhile, the estimation algorithm accounts for 3% of the published articles on software defect prediction. Association methods have 2%. Clustering methods have four total published articles, with 4%. Finally, data processing has 14 published articles, resulting in 13%. The distribution of datasets used in software defect prediction found that 21 published articles utilised private datasets, a result of about 40%. Public datasets have a total of 32 published articles on software defect prediction, which resulted in 60%. Additionally, it was found that Naïve Bayes dominates the most...
Published
How to Cite
Issue
Section
FUDMA Journal of Sciences