Partner Recommendation Based on Scholar Embedding Method

The work recommended by research collaborators provides scholars with a wealth of collaborative suggestions, expands their societal relationship, and accelerates the output of research results. It is difficult for general recommendation algorithms to handle social relationship networks with large node sizes and complex and dense cooperative relationships. In this paper, we propose a scholar embedding recommendation model (SERM), which uses scholars' academic ability and research interest similarity to weigh the edges of the cooperative network on the basis of scholars' cooperative network, and uses a random walk algorithm to sample cooperative paths on the weighted network, and then uses node embedding technology to embed scholars. Finally, the top k scholars in the same cluster as the target scholars are recommended using a clustering algorithm. The experiments show that the model is able to effectively mine the potential collaborations of scholars, and the recommended scholars match the target scholars to a high degree. Compared to other models, the precision, recall, and F1 metrics of recommendations improved by 2.2%, 2.6%, and 2.4 %.