Analysis of duplicate issue reports for issue tracking system

Issue tracking is a core part of the software development process. For open source projects, the number of duplicate reports represents a significant percentage of the repository, numbering in the thousands of reports for popular projects. In this paper, we introduce an approach to suggest potential duplicate issue reports to the issue reporter who is submitting a new report. We have evaluated the accuracy of our approach analytically against the Firefox, Eclipse platform, Apache and Mylyn projects, achieving a range of 66%-100% recall rate on reports from the four projects issue repositories. The recall rates are similar to others reported but we remove restrictions or constraints on previous approaches. With the promising recall rate, we are looking at the possibility to integrating this approach with commercial issue tracking software and ticketing systems. The initial implementation shows that its use reduces the duplicate issue reports and therefore improves the efficiency to process the unique and important issues.