A hybrid approach for developer recommendation based on social network

Abstract

Bug fixing requires collaboration among developers, however, most of current studies recommend developers based on the textual content and similarity of bug reports, ignoring the implicit social relationships generated by developers when they collaborate on fixing tasks. This paper proposes a hybrid approach for bug developer recommendation to address this problem. The method constructs a developer social network, extracts developer social relationships implicitly modeled in bug report history records and comment, and uses Graph Convolutional Neural Network (GCN) to learn the closeness between developers. In addition, Convolutional Neural Networks (CNN) are used to learn the bug report text content, suitable developers are recommended by fusing the collaborative relationships between developers and the textual content of bug reports. According to experimental findings, the method's accuracy is significantly better than that of other methods.