Software defect prediction using transfer method

Traditional machine learning works well within company defect prediction. Unlike these works, we consider the scenario where source and target data are drawn from different companies, recently referred to as cross-company defect prediction. In this paper, we proposed a novel algorithm based on transfer method, called Transfer Naive Bayes (TNB). Our solution transferred the information of test data to the weights of the training data. The theoretical analysis and experiment results indicate that our algorithm is able to get more accurate result within less runtime cost than the state of the art algorithm.