A Fusion Information Embedding Method for User Identity Matching Across Social Networks

Yizhuo Yang, Hongtao Yu, Ruiyang Huang, Tuosiyu Ming

Abstract: Aiming at the deficiencies of low accuracy and difficulty in obtaining data for existing cross social network user identity matching algorithms, an algorithm of user identity matching across social networks based on information fusion representation was proposed. Firstly, the seed nodes were used to transform the cross-network problem into a single network problem by using the network merging algorithms. Then the username information was turned to vectors and merged with the topology vector. Finally, with the network representation learning method, account nodes' vectors with information of usernames and topology were acquired for the mission of user identity matching. Experimental results showed that the average F1 measure reached 79.7%, which is improved by 7.3%-28.8% compared with traditional machine learning algorithms and the existing other two algorithms. It can be seen that our algorithm can effectively improve the performance of user identity matching.