Network embedding based link prediction in dynamic networks

Link prediction is a fundamental task in network theory due to the wide variety of applications in different domains. The objective of link prediction is to find the future links that are likely to be seen in some future time. In this paper, we propose a novel embedding-based technique that utilizes the concept of the Skip-gram framework. An embedding-based method embodies the learning of feature representations of nodes or links in a network. Our method jointly exploits the Skip-gram framework and max aggregator for edge embedding tasks. To test the effectiveness of the proposed method, we have conducted experiments on large size real-world networks. In the experimental evaluation, we have compared the proposed method against both similarity-based and learning-based approaches. The experimental results indicate the effectiveness of the proposed method both in terms of time and accuracy. (C) 2021 Elsevier B.V. All rights reserved.