"TCP Transfer Learning Feature Vectors: 270,611 Labeled Test Execution Records Across 17 Defects4J Projects"
收藏DataCite Commons2026-04-02 更新2026-05-03 收录
下载链接：
https://ieee-dataport.org/documents/tcp-transfer-learning-feature-vectors-270611-labeled-test-execution-records-across-17
下载链接
链接失效反馈官方服务：
资源简介：
"This dataset comprises 270,611 labeled test execution feature vectors extracted from 17 Java projects in the Defects4J benchmark (version 3.0), spanning 915 real-world bugs. Each record corresponds to a single test case execution and includes 13 features categorized as History (HIS) features \u2014 failure rate, last failure, and recent failure counts over fixed windows \u2014 and Coverage (COV) features \u2014 line, branch, and condition coverage percentages and absolute counts extracted from Cobertura XML reports. The binary label field indicates whether the test case revealed a fault (1) or passed (0). Data is provided as one CSV file per project (17 files total). The dataset was constructed to support research on cross-project transfer learning for TCP under cold-start conditions, where a target project lacks sufficient historical execution data to train a reliable within-project classifier."
提供机构：
IEEE DataPort
创建时间：
2026-04-02
