Enhanced Ensemble Prediction Algorithms for Detecting Faulty Modules in Object Oriented Systems Using Quality Metrics

The high usage of software system poses high quality demand from users, which results in increased software complexity. To address these complexities, software quality engineering methods should be updated accordingly and enhance their quality assuring methods. Fault prediction, a sub-task of SQE, is designed to solve this issue and provide a strategy to identify faulty parts of a program, so that the testing process can concentrate only on those regions. This will improve the testing process and indirectly help to reduce development life cycle, project risks, resource and infrastructure costs. Measuring quality using software metrics for fault identification is gaining wide interest in software industry as they help to reduce time and cost. Existing system use either traditional simple metrics or object oriented metrics during fault detection combined with single classifier prediction system. This study combines the use of simple and object oriented metrics and uses a multiple classifier prediction system to identify module faults. In this study, a total of 20 metrics combining both traditional and OO metrics are used for fault detection. To analyze the performance of these metrics on fault module detection, the study proposes the use of ensemble classifiers that uses three frequently used classifiers, Back Propagation Neural Network (BPNN), Support Vector Machine (SVM) and K-Nearest Neighbour (KNN). A novel classifier aggregation method is proposed to combine the classification results. Four methods, Sequential Selection, Random Selection with No Replacement, Selection with Bagging and Selection with Boosting, are used to generate different variants of input dataset. The three classifiers were grouped together as 2-classifier and 3-classifier prediction ensemble models. A total of 16 ensemble models were proposed for fault prediction. The performance of the proposed prediciton models was analyzed using accuracy, precision, recall and F-measure. When comparing with single classifier systems all the proposed models produced improved classification performance and among the 16 multiple classifier models, the 3-classifier model that combined BPNN, SVM and KNN produced best results. Prediction of software module defection can be improved by combining simple and object oriented metrics with multiple classifiers.