Name disambiguation, which aims to distinguish real-life person from documents associated with a same reference by partition the documents, has received extensive concern in many intelligent tasks, e.g., information retrieval, bibliographic data analysis and mining system. Existing methods implement name disambiguation utilizing linkage information or biographical feature, however, only a few work try to combine them effectively. In this paper, we propose a novel model that incorporates structural information and attribute features based on the Graph Convolutional Network to learn discriminating embedding, and achieves individual distinction by equipping a hierarchical clustering algorithm. We evaluate the proposed model on real-world academic networks Aminer, and experimental results show that the proposed method is competitive with the state-of-the-art methods.