Heterogeneous cross-project defect prediction with multiple source projects based on transfer learning

Abstract

Cross-project defect prediction (CPDP) aims to predict the defect proneness of target project with the defect data of source project. Existing CPDP methods are based on the assumption that source and target projects should have the same metrics. Heterogeneous cross-project defect prediction (HCPDP) builds a prediction model using heterogeneous source and target projects. Existing HCPDP methods just focus on one source project or multiple source projects with the same metrics. These methods limit the scope of getting the source project. In this paper, we propose Heterogeneous Defect Prediction with Multiple source projects (HDPM) which can use multiple heterogeneous source projects for defect prediction. HDPM based on transfer learning which can learn knowledge from one domain and use it to help with other domain. HDPM constructs a projective matrix between heterogeneous source and target projects to make the distributions of source and target projects similar. We conduct experiments on 14 projects from four public datasets and the results show that HDPM can achieve better performance compared with existing CPDP methods, and outperforms or is comparable to within-project defect prediction method. The use of multiple heterogeneous source projects for defect prediction can effectively extend the data acquisition range of defect prediction and make software defect prediction better applied to software engineering.