الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 185 |  |  |
| | | from | 185 | | |
Abstract
Recently, research studies have been directed to the construction of a universal defect prediction model. Such models are trained using different projects to have enough training data and be generic. One of the main challenges in the construction of a universal model is the different distributions of each software metric among various projects. This study aims to build a universal defect prediction model to predict software defective classes. It also aims to validate the Object-Oriented Cognitive Complexity metrics suite (CC metrics) for its association with fault-proneness. Finally, this study aims to compare the prediction performances of each of the CC metrics, the Chidamber and Kemerer metrics suite (CK metrics), and a combination of both suites, taking into account the effect of preprocessing techniques applied to them. A neural network model is constructed using three object-oriented metrics sets: the CK metrics, the CC metrics, and a combination of both. Different preprocessing techniques are applied to these metrics sets to overcome the variations in their distributions among various projects. The CK metrics perform well whether a preprocessing technique is applied or not, while the CC metrics’ performance is significantly affected by different preprocessing techniques. The CC metrics always outperform in the recall, while the CK metrics usually outperform in the total accuracy, AUC of ROC, precision, F-measure, and MCC. The combination of both the CK and CC metrics exhibits a balance between different performance metrics rather than a superiority in a certain performance metric with a large difference from others. Both quantization and quantization with normalization preprocessing techniques have very close performance. Normalization preprocessing results in the highest recall values using different metrics sets compared to other preprocessing techniques. In conclusion, the construction of a universal model is applicable using different preprocessing techniques and different object-oriented metrics suites. The CC metrics are validated for their association with software fault-proneness. Preprocessing improves the prediction performance when applied to the CC metrics, but it has minimal effect on the prediction performance when applied to the CK metrics.