User Alignment across Dynamic Social Networks based on Heuristic Algorithm

Abstract

User alignment across social networks, whose main goal is to fuse user information in different network platforms, is a fundamental task in social network analysis. It can benefit social network applications such as user recommendation and information diffusion. Attributed to the inherent dynamic characteristic of the social networks, aligning users in dynamic networks is a key issue in practice. However, most of the alignment models encounter model retraining when the network is updated, thus result in the consumption of time and resources. To address this problem, a heuristic algorithm is proposed to align users in a dynamic environment. Firstly, the attention mechanism is leveraged to obtain the local importance weight of the new node in a single network. Secondly, the anchor nodes are adopted as supervised information for heuristically learning the alignment task-driven local influence of new nodes. Finally, by preserving the second-order similarity of the network, the model aligns users across networks. Experimental results conducted on realworld datasets prove that the proposed model has a comparable performance but lower time complexity compared with several state-of-the-art algorithms.