الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 62 |  |  |
| | | from | 62 | | |
Abstract
One of the crucial attributes of the bug report is severity. Accurate prediction of bug severity can be a great contribution towards optimized software maintenance. As software systems become very large-scaled and more complicated, manual classification for bugs does not longer meet the need for quick and accurate classification. Solving minor bugs can be considered as a luxury for most of software companies. Accurate classification will lead to solving the blocking bugs first by giving it the highest priority, hence the quality and reliability of software system are improved, in addition to saving time and expense. Automated classification is not only more accurate than manual classification but also faster. So that the need for automated classification for bugs is urgently needed. In this research, a new approach for classifying bugs is described. This new approach depends on ensemble of classification techniques, that’s why it is with better accuracy than previously presented approaches in this field. Ensemble of classifiers is more accurate than any of its individual base classifier with two necessary conditions, firstly, the classifiers must be accurate, and secondly, the classifiers must be diverse. The proposed technique combines ensemble of different classification techniques to increase the accuracy of detecting the severity of a bug. The chosen classification techniques are combined based on Customized Cascading Weighted Majority Voting. The proposed technique has been evaluated using datasets from open-source projects. Experiments show that the proposed technique has superior performance compared to other classification techniques especially when the used datasets have imbalanced class distribution.