Authors:
Afrina Khatun
and
Kazi Sakib
Affiliation:
University of Dhaka, Bangladesh
Keyword(s):
Bug Assignment, Bug Report, Commit Log, Tf-idf Weighting.
Abstract:
Automatic bug reports assignment to fixers is an important activity for software quality assurance. Existing
approaches consider either the bug fixing or source commit activities which may result in inactive or inexperienced developers suggestions. Considering only one of the information can not compensate another leading
to reduced accuracy of developer suggestion. An approach named BSBA is proposed, which combines the
expertise and recency of both bug fixing and source commit activities. Firstly, BSBA collects the source code
and commit logs to construct an index, mapping the source entities with their commit time, which presents
developers’ source code activities. Secondly, it processes the fixed bug reports to build another index which
connects the report keywords to the fixing time. Finally, on arrival of new reports, BSBA queries the two
indexes, and combines the query results using tf-idf weighting technique to measure a BSBA score for the
developers. The top scored de
velopers are suggested as appropriate fixers. BSBA is applied on two open
source projects - Eclipse JDT and SWT, and is compared with three existing techniques. The results show
that BSBA obtains the actual fixer at Top 1 position in 45.67%, and 47.50% cases for Eclipse JDT and SWT
respectively, which is higher than the existing approaches. It also shows that BSBA improves the accuracy of
existing techniques on average by 3%-60%.
(More)