Attributed Heterogeneous Network Embedding Based on Graph Convolutional Neural Network

Network embedding transforms a network into a low-dimensional space, in which the information of network can be stored, it is an effective approach to solve the problems of network analysis. However, most of the networks in real scenario are attributed heterogeneous network (AHN), which contains multiple types of nodes and edges. In addition, different types of nodes in AHN have different attribute information. To deal with the heterogeneity of topological structure and node attributes of AHN while preserving the rich information of AHN as much as possible in embedding space, is main difficulty in AHN embedding. To solve this problem, this paper proposes a novel AHN embedding model, namely AHN2Vec (Attributed heterogeneous network to vector), which can preserve node attributes, k-order neighborhood attributes, semantic information and topological structure information in embedding space. Experimental results on two real-word datasets not only show that the overall performance of AHN2Vec is better than the state-of-the-art methods, but also demonstrate AHN2Vec is able to generate high-quality node embeddings for AHN and can achieve excellent performance in downstream tasks.