An Automated approach for Bug Categorization using Fuzzy Logic

Abstract

Various automated techniques built to benefit software developers, bug triagers, stakeholders and users in open source systems, utilize information placed in issue tracking systems. The success of these techniques depends largely on the quality of information present in the issue reports. Assigning correct label to issue reports is one of the quality concerns. Previous empirical studies conducted on the issue reports show that most issues are either mislabeled or are not labeled at all. Thus, in order to enhance quality of issue reports, there is a strong need to propose an automated and accurate bug labeling approach. A label can be a bug, feature enhancement or other request. In this paper, we propose an automated approach to label an issue either as bug or other request based on fuzzy set theory. Experiments are conducted on issue repository of three open source software systems: HTTPClient, Jackrabbit and Lucene. We have achieved an accuracy of 87%, 83.5% and 90.8% and F-Measure score of 0.83, 0.79 and 0.84 respectively. This is a considerable improvement as compared to the earlier reported work on these three datasets using topic modeling approach.