Exploring Network Embedding for Efficient Message Routing in Opportunistic Mobile Social Networks

Abstract

With the advancement in communication technologies and the widespread availability of mobile devices, the opportunistic mobile social networks (OMSNs) are gaining momentum in supporting spontaneous communication and interaction among end-users who opportunistically contact each other. However, existing research on message routing in OMSNs face major challenges on achieving a high routing efficiency and low latency for social information request. This paper proposes a personalised message routing (PMR) framework that leverages an inductive network embedding model and an attention-based mechanism to facilitate efficient message routing in opportunistic networks. Specifically, the network embedding model encompasses a higher-order proximity profiling algorithm in order to embed both the content-based and structure-based network features beyond immediate friends into low dimensional representations. Further, we present an attentional neural network model to learn user-friend preferences, for the purpose of capturing the diversified interests among connected users and to determine the most informative friends during the message dissemination process. The performance of our proposed framework is evaluated through simulations on three real-world mobile network trace datasets and the experimental results show that the proposed PMR framework considerably and consistently outperforms the state-of-the-art message routing methods.