Abstract
Open source software projects such as Mozilla and Eclipse have a huge number of bug reports submitted by their users who are distributed all over the world. Handling these reports and assigning a relevant developer to fix them have been performed manually, which is costly in terms of time, effort, and operation. In this paper, an automated bug triaging system is examined using a hybrid machine learning technique and a novel feature selection method. This model was evaluated using Mozilla Firefox projects with respect to accuracy, precision, recall and F-measure metrics. The collected data set consists of 65 products from 1991 to 2016, including 542 components, 135490 bug reports and 807 developers; these bug reports were distributed over 10 datasets. The model is examined using bagging, boosting, and decorate ensemble methods along with Bayes Net, Naive Bayes, Decision Table, Random Tree and J48 base learner classifiers. The results illustrate that decorate and bagging ensemble methods have the ability to improve the classification results, which eventually leads to improve the maintenance process. The experiment results have achieved a recall ratio of up to 96%.