Cross-Project Defect Prediction Using a Credibility Theory Based Naive Bayes Classifier.
Abstract
Several defect prediction models proposed are effective when historical datasets are available. Defect prediction becomes difficult when no historical data exist. Cross-project defect prediction (CPDP), which uses projects from other sources/companies to predict the defects in the target projects proposed in recent studies has shown promising results. However, the performance of most CPDP approaches are still beyond satisfactory mainly due to distribution mismatch between the source and target projects. In this study, a credibility theory based Naïve Bayes (CNB) classifier is proposed to establish a novel reweighting mechanism between the source projects and target projects so that the source data could simultaneously adapt to the target data distribution and retain its own pattern. Our experimental results show that the feasibility of the novel algorithm design and demonstrate the significant improvement in terms of the performance metrics considered achieved by CNB over other CPDP approaches. Several transfer leaning techniques [12], [13], [26] have been proposed to tackle the data distribution differences. These techniques extract knowledge from a source of data and transfers it to another data. The transferred knowledge is employed during model construction. Keywords-cross-project defect prediction; credibility theory; Naïve Bayes classifier; transfer learning; software engineering; quality assurance I. Considering the inferior performance in cross-project defect prediction and on the basis of transfer learning, we seek to use credibility theory to narrow the difference in feature distribution among datasets to obtain better crossproject prediction performance. This study adopts the credibility theory technique from the field of actuarial science generally used for insurance products pricing. The theory helps an insurer to set reasonable premium price for the insured according to the limited data from other insured and collected data (even less relevant) from its own insurance company [7]. Cross-project software defect prediction could be improved by the heuristic application of credibility theory in insurance product pricing. In the study, the small-sc