Multi-Path Relationship Preserved Social Network Embedding

Abstract

Social network embedding, namely, embedding social network nodes into a low-dimensional space, is the foundation of social network analysis, such as node classification and link prediction. Although many existing methods attempt to address this task, most of them only consider the shallow relationship between two nodes in the network, which ignore capturing multiple and semantic-rich social relationships between users. To this end, we define such multiple and semantic-rich relationships as multi-path relationships, and propose a multi-path relationship preserved social network embedding method named MPR-SNE, which is based on the recurrent neural network framework that incorporates both social network structure and node profile information. Specifically, we first utilize random walks to explore the multiple social relationship paths between nodes. Then, a new recurrent unit called bi-directional multi-path relationship unit is proposed to better capture the properties of multi-path relationships. Finally, two objective functions are designed to seamlessly integrate social network structure and node profile information into node representation. The experimental results on two real-world networks show that MPR-SNE outperforms the state-of-the-art baselines on node classification task and link prediction task.

Keywords

• Social network embedding
• multi-path relationship
• node profile information
• RNN