Predicting fault-prone software components in telecommunication assembly patches with imbalanced datasets

Software quality is imperative in the telecommunication industry. This has led to research towards the development of classification models, able to give a pre-release prediction of whether a software module is faulty or not. However, the data collected within this domain are typically imbalanced. This causes bias on the performance of any prediction model. In this study a hybrid framework for evaluating the fault-proneness of telecommunication software components in a highly imbalanced setting is presented.