Recommending Bug Assignment Approaches for Individual Bug Reports: An Empirical Investigation

Multiple approaches have been proposed to automatically recommend potential developers who can address bug reports. These approaches are typically designed to work for any bug report submitted to any software project. However, we conjecture that these approaches may not work equally well for all the reports in a project. We conducted an empirical study to validate this conjecture, using three bug assignment approaches applied on 2,249 bug reports from two open source systems. We found empirical evidence that validates our conjecture, which led us to explore the idea of identifying and applying the best-performing approach for each bug report to obtain more accurate developer recommendations. We conducted an additional study to assess the feasibility of this idea using machine learning. While we found a wide margin of accuracy improvement for this approach, it is far from achieving the maximum possible improvement and performs comparably to baseline approaches. We discuss potential reasons for these results and conjecture that the assignment approaches may not capture important information about the bug assignment process that developers perform in practice. The results warrant future research in understanding how developers assign bug reports and improving automated bug report assignment