ReDef (Revert-based Defect dataset)
收藏资源简介:
ReDef数据集是从22个大规模的C/C++项目中收集的,它是一个高置信度的即时软件缺陷预测数据集,其中包含通过回滚操作锚定和GPT辅助过滤的函数级修改。该数据集进一步通过手动抽样验证其可靠性,包含3,164个缺陷修改和10,268个清洁修改,总共有13,332个条目。ReDef数据集旨在为即时软件缺陷预测提供更可靠的数据支持,帮助研究人员更好地理解和预测代码中的缺陷。
The ReDef dataset is collected from 22 large-scale C/C++ projects. It is a high-confidence dataset for just-in-time software defect prediction, containing function-level changes anchored via rollback operations and filtered with GPT assistance. Its reliability was further validated through manual sampling, and it includes 3,164 defect-inducing changes and 10,268 clean changes, totaling 13,332 entries. The ReDef dataset aims to provide more reliable data support for just-in-time software defect prediction, helping researchers better understand and predict defects in code.




