الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 102 |  |  |
| | | from | 102 | | |
Abstract
Mining software repositories is a growing research field where rich data, available inthe different development software repositories, are analyzed and cross-linked to uncoveruseful information. This useful information helps in answering many important softwarequestions such as: is the introduced class/ change bug prone or not?, how severe the bugis? and what test case(s) was used to find a defect?The core idea of this thesis is bug detection. Detecting bugs as early as they are introducedto the version control system would definitely help in saving time and effortexerted on discovering bulks of hidden old bugs in the system during testing or maintenancephases.A new defect prediction technique is proposed. It focuses on both wings of any predictiontechnique, selected metrics and classifier techniques. The main contribution is combiningchange metrics with ensemble classifiers in order to reach better results. Change metricsare process metrics that capture changes done across different releases of a certain project.Experiments showed that using change metrics with ensemble classification techniquesproduced the best results. Change metrics outperformed static code metrics and thecombined model of change and static code metrics. Ensembles tend to be more accuratethan their base classifiers. Moreover, they were empirically observed to solve the classimbalance problem.Following the achieved results, it can be concluded that building defect prediction modelsusing change metrics and ensemble classifiers is highly recommended, especially when thedatasets used have imbalanced class distribution.