Automatic Software Bug Triage Method Based On Multi-feature Bug Tossing Graphs

Problem plagued users and software developers-software defects, not only increase the cost ofthe software organization, endanger the quality of software products, but also slow down theprogress of the software development and release. Timely identification and correct software defectof software engineering practices are very important. Large software projects are using Bugzillatool to manage bug reports. The triagers filter duplicate bugs and assign the non duplicate bug todevelopers, however, most bugs are assigned manually to developers. Such as the Eclipse andMozilla, received hundreds of bug reports per day, and then assigned each of bugs to one of thethousands developers to repair, and more than half of the bugs will appear bug tossing, which willgreatly reduce the bug repair efficiency and improve the bug repair time. Accurately and quicklyidentifying duplicate bug reports and assigning the bug to the right developer, and predicting thebug repair time is urgent need problems.(1) The method uses multi-feature classifier to detect the duplicate bug, and all the bug reports inthe repository are organized into a bucket structure.(2) The method is used in order to forecast the developer of the bugs based on multi-feature bugtossing graphs and classifier based on latent semantic indexing. This paper extracted therelevant features like report title, report summary etc., from each bug report, and extracteddeveloper’s name who resolved the bug reports from the developers activity, obtained theterm-to-document matrix using parsing, filtering and term weighting methods. This paperreduced the dimensionality of the obtained term-to-document matrix by applying featureselection and latent semantic indexing methods. Combining with tossing transactionprobabilities between developers, we build the tossing graphs and attach the bug class (productand component) to tossing graphs. According to the classifier, we can predict the potentialdevelopers list for bugs, combining with multi-feature bug tossing graph updating thedevelopers.(3) A weighted similarity model is suggested to match the summary, description and comments for apair of software bugs. Then the fix duration of bug is the average fix time of all the similar bugs.