Academic networks in the real world can usually be described by heterogeneous information networks composed of multi-type nodes and relationships. Some existing research on representation learning for homogeneous information networks lacks the ability to explore heterogeneous information networks in heterogeneous information networks. It cannot be applied to heterogeneous information networks. Aiming at the practical needs of effectively identifying and discovering scientific research teams from the academic heterogeneous information network composed of massive and complex scientific and technological big data, this paper proposes a scientific research team identification method based on representation learning of academic heterogeneous information networks. The attention mechanism at node level and meta-path level learns low-dimensional, dense and real-valued vector representations on the basis of retaining the rich topological information of nodes in the network and the semantic information based on meta-paths, and realizes effective identification and discovery of scientific research teams and important team members in academic heterogeneous information networks based on maximizing node influence. Experimental results show that our proposed method outperforms the comparative methods.