[Purpose/significance] Multiple networks is learned by network embedding to discover the relevance be?
tween authors and recommend collaborators better. [Method/process] First, collect relevant literature data in the field
of information science as the original data set. After data cleaning, multiple academic information networks of each in?
formation entity are constructed according to the multiple relationships between authors. Then use the Node 2vec to
learn the information of each node for the high-dimensional network, so as to obtain the vector representation of the
nodes in each network. Secondly, the author similarity in each network is calculated by cosine similarity. Finally, the fi?
nal recommendation result is obtained by integrating the institutional cooperation preference between authors and the
similarity of the authors' academic level. [Result/conclusion] The fusion model proposed in this paper considers multi ?
ple networks and data sparsity, and performs better than a single dimension in AUC value, and obtains a better predic?
tion effect for collaborators. The experimental results show that the collaborator recommendation model is feasible in
the author collaborator recommendation in the field of information science.