Cross-Network User Matching Based on Association Strength

Cross-network user matching is the one of the fundamental problems in social network fusion and analysis. This paper proposes an unsupervised algorithm based on association strength to address this problem. Specifically, the algorithm firstly computes the association strength between nodes based on user behavior characteristics in the social network graph. Then, the node sequence is obtained by the biased walk method based on the association strength, and word2vec method is utilized to learn the node feature vector. Finally, cross-network user matching is achieved based on the distance of feature vectors. The highlight of this paper is that the bias walking based on the association strength can select nodes as the context to form a specific sequence, and then do benefits on feature vectors learning. The experimental results show that the proposed algorithm can effectively match user without any prior knowledge. Furthermore, the matching precision is improved compared to the baseline method.