Towards evolving software recommendation with time-sliced social and behavioral information

Hongqi Chen, Zhiyong Feng, Shizhan Chen, Xiao Xue, Hongyue Wu, Yingchao Sun, Yanwei Xu, Gaoyong Han

Abstract: Software recommendations play a crucial role in helping developers discover potential functional requirements and improve development efficiencies. As new requirements emerge in the software development process, developers’ preferences tend to change over time and social relationships. However, the existing works fall short of capturing the evolution of developers’ interests. To overcome these problems, evolving software recommendation with time-sliced social and behavioral information is proposed for capturing the dynamic interests of developers. Specifically, the different behaviors of developers are considered and graph structure features on projects are extracted by gated graph neural networks. Then, the graph attention networks are introduced to model rich developer-project interactions and social aggregation. Finally, the integration of time-sliced representations on the developer and project sides is employed through gated recurrent units to capture the dynamic interests of developers. Extensive experiments conducted on three datasets demonstrate the superiority of the proposed model over representative baseline methods across various evaluation metrics.