Assignment of bug reports to software developers using a multi-population evolutionary method

Existing approaches assign bug reports using only data from previously fixed reports. This can result in assignments to inactive developers, as well as not considering newcomers. A significant portion of assignments typically do not consider the workload of developers, which can overwhelm some and make the revision/debugging/correction process more time-consuming. This work proposes an approach for assigning bug reports that combines the experience and recent activities of developers, as well as their workload. When a new report is received, the effort to fix the bug based on similar error is estimated and each developer's affinity with the file containing the bug is calculated using a Fuzzy Inference system. Subsequently, the Golden Ball, a multi-population evolutionary metaheuristic, is used to assign these reports to developers according to affinity and workload. Experimental results show that, when compared with a brute force algorithm, the proposed approach reaches optimal values for assign in most cases (75% of the analyzed scenarios). The approach also obtained significantly better averages in 92.30% of the instances when compared to a Genetic Algorithm and 84.61% when compared to a Distributed Genetic Algorithm, and in only 23.07% of the instances there was no significant difference between the techniques.