Multi-source cross-project defect prediction method by marking defect severity

Software defect prediction is a significant measure to ensure software quality. Because there are no sufficient historical data within a project to train the classifier, cross-project defect prediction (CPDP) has been recognized as a fundamental approach. A two-phase transfer learning model (TPTL) for CPDP is proposed to address the limitation of TCA+. However, its algorithm time complexity is too high, and it cannot make full use of sufficient source projects. In this paper, we propose a multi-source cross-project software defect prediction method for defect severity marking (MSDSM). MSDSM first builds a multi-source model to fully take advantage of knowledge recorded in all candidate source projects and preprocesses metric value to make feature distributions similar. Defect severity marking (DSM) quantitatively describes defect severity for each instance. Experiment results show that, on average across 42 datasets, MSDSM respectively improves TCA+ and LT by 3.71% and 25.66% in terms of F1score; and LT, TCA+, and TPTL by 88.35%, 9.60%, and 6.01% in terms of cost effectiveness. Moreover, its time complexity is lowest due to its simple structure.