Abstract: Bug reports are an integral part of the software development process. They are used by software developers to improve the quality of the software. Bug triaging deals with the selection of a suitable developer to resolve the bug. With the increase in the number of bugs, this process only becomes troublesome and laborious. If the bug is assigned to a developer who is not able to resolve the bug, it is reassigned to another developer. This bug tossing leads to delays in resolving the bug and a lot of wasted resources. The selected bugs are then prioritized based on their severity and fixed according to the priority. If the severity is incorrectly reported, it results in a waste of time and resources.
In this paper, we present how classical machine learning classification algorithms can be used in bug tracking systems during the process of bug reporting to suggest the severity of the bug and developers(assignee) using NLP (Natural Language Processing) techniques on the summary of the bug report. The predictions from these classification algorithms are then incorporated in the bug report filing and assignment phase of the bug life cycle.
We have collected bug reports from Bugzilla for two open-source projects: Eclipse and LibreOffice and compared the results of various classification algorithms. Even though fully automated assignment is not present, the prediction accuracies are high enough to be used as suggestions to the reporter/assigner in our bug tracking system.

Keywords: Severity Prediction, Bug Triaging, Project management, text-based classification, NLP, TF-IDF.

PDF | DOI: 10.17148/IJARCCE.2022.11825

Open chat
Chat with IJARCCE