BICON:Connecting the Same Users of Different Social Networks using BiLSTM

Nowadays, there are many online social media. People often have different accounts in different social platforms to enjoy the specific services provided by different online social media, and connecting the same users in different social networks is beneficial to personalized service recommendation, social network data mining, etc. This paper proposes a method based on BiLSTM model to connect the same users in different social network called BICON, it is mainly divided into two parts. The social network graph sampling part: sampling the network, encoding all nodes in the network into vector representations, capturing local and global network structure information, similar users will have similar representations in the embedded space. Network connection part: By constructing a deep network model, input vector representations of different social networks, train models according to labeled data, align the anchor nodes therein, map all users of one network to another network, and then judge by user embedded vector similarity whether the two represent the same user. BICON leverages the ability of deep learning to extract features automatically and consistently models users across different social networks. A lot of experiments have been carried out on two real social networks. The results show that the proposed method can effectively align anchor nodes according to network structure information, and BICON is superior to the most advanced method in link accuracy.