Cross-Project Defect Prediction Based on Dynamic Distributed Adaptive Networks
In software defect prediction,cross-project defect prediction is based on training models by using the labeled data from the original project,and predicts the defects of the current development target project.However,the data distribution differences be-tween two different projects often limit the ability of cross-project defect prediction models.As the data from the source domain and target domain usually come from different distributions,existing methods mainly adapt to cross-domain edges or conditional distribu-tions.In practical application,existing methods are unable to quantitatively evaluate the importance of marginal and conditional distri-butions,which leads to unsatisfactory transmission performance.This paper proposes a cross-project defect prediction method based on a dynamic distribution adaptation network to address the distribution difference,the transmission learning is used to quantitatively evaluate the relative importance of each distribution.the proposed method is verified by the experiments on 24 projects from 3 public datasets.The results show that,on average,the proposed method is superior to all baseline methods,the proposed method reaches the AUC and F1 scores by at least 1.3％and 5.7％,respectively.This indicates that the proposed method has good performance char-acteristics.
transfer learningcross-projectdynamic distributiondeep learningquantitative evaluation
NETL
NSTL
万方数据
