A bug assignment technique based on bug fixing expertise and source commit recency of developers
Afrina Khatun, Kazi Muheymin Sakib · 2016
Automatic bug assignment is an essential activity aiming at assigning bugs to appropriate developers. Existing approaches consider either recent commits or previous bug fixes of developers, leading to recommendation of inexperienced or inactive developers respectively. Considering only one information source leads these approaches to low prediction accuracy. An approach called ERBA is proposed, which considers both expertise and recent activities of developers. ERBA first processes source code and commit logs to construct an index connecting the source entities with developer recent activities. Next, it takes fixed bug reports and builds another index, mapping the bug report keywords with developer bug fixing expertise. On arrival of new bug reports, the final module queries the two indexes using the new bug report terms, and applies tf-idf technique on the query result to calculate an ERBA score for developers. Finally, an ascending ordered list on ERBA score is suggested. For assessment of competency, a case study has been conducted on Eclipse JDT. It depicts that ERBA outperforms existing approach by improving prediction accuracy from 33.8% upto 44%. The result also represents that ERBA shows the first correct developer on average near 4.04 ranks, whereas existing approach shows in 7.27.