DENATURE: duplicate detection and type identification in open source bug repositories

Software projects reckon on the bug tracking systems to guide software maintenance activities. The critical information about the nature of the crash is carried by the bug reports which are submitted to bug repositories. This information is in free form text format and is submitted by users or developers. A large amount of bug reports gets collected in bug repositories. Out of these submitted bugs, many reports are mere identical of the already existing bugs. Furthermore, not all non-duplicate bugs are reproducible in nature. This paper introduces DENATURE, a two step framework for detecting duplication and identifying bug type. The proposed framework will help to minimize time and developer's effort utilized in resolution of bug reports which will further improvise overall software quality. Information retrieval techniques are used for finding duplicate bugs and machine learning classification techniques are used for identifying the type of bug report. Through experiments, we found that the proposed framework obtained prediction accuracy up to 88.81%.