Comprehensive Bibliographic Survey and Forward-Looking Recommendations for Software Defect Prediction: Datasets, Validation Methodologies, Prediction Approaches, and Tools

The development of reliable software depends heavily on the effective collaboration between teams responsible for development and testing. Despite ongoing efforts, many software programs still contain bugs that can lead to financial losses and business risks. Therefore, detecting and fixing software defects after release is crucial. While binary classification methods have been commonly used for this purpose, recent Artificial Intelligence (AI) advancements offer new opportunities for software teams to create more robust software. To address challenges in Software Defect Prediction (SDP), we conducted a thorough bibliographic survey of 79 research articles from the year 2011 to 2023 that examined previous models, datasets, data validation techniques, defect detection, prediction methods, and SDP tools. The survey revealed that previous research often lacked appropriate datasets with the necessary characteristics and data validation methods. Additionally, many standard datasets suffer from a lack of labels, which hinders effective defect detection. Systematic literature reviews on SDP are scarce, further emphasizing the importance of this study. Based on the findings, we provide crucial recommendations for designing effective SDP models and tools. The proposed survey outlines an architecture for constructing SDP datasets with the appropriate characteristics, as well as multi-label classification and data validation methodologies for software defects. This approach aims to enhance SDP research and contribute to the development of high-quality software products by improving defect prediction accuracy.