Learning Advisor-Advisee Relationship from Multiplex Network Structure

The analysis of the advisor-advisor relationship can provide rich information for understanding the interactions among entities in academic network. Most of the existing methods mine the relationship in collaboration network. They focus on extracting deep relationships between attributes, but ignore the structure of nodes. Compared with the multiplex network, the collaboration network can only provide a single view, which describes whether the two nodes have co-authored publications. To tackle these problems, we propose an Attention-based Multiplex Network Structure Fusion (AMNSF) method for mining the advisor-advisee relationship. AMNSF takes the multiplex network as input composed of the acknowledgment network and the collaboration network where the acknowledgment network provides the social relationship between entities. We make full use of the network’s structure information and design a novel network fusion mechanism to integrate the structure information from each layer. Finally, we conduct extensive experiments on a variety of datasets. The experimental results show that the proposed model outperforms the state-of-the-art methods.