A New Algorithm for Social Inference Using Position Information

With the development of relationship inference technology, the available social data increases with the popularity of Internet and information technology. This paper focuses on the social relation, which is one of the most sensitive information that can be inferred from the association between location data and user data. Besides, the paper proposes a new random walk method based on node weight, which relies on PageRank(PR) to reallocate node weight and limit the range of random walk to improve the accuracy of prediction. Compared with the existing optimal method, the proposed inference method uses SkipGram neural network model to embed the graph. Finally, a holistic relationship mining algorithm based on node weight is proposed. The advantages of this algorithm can be effectively applied to scenarios with high accuracy requirements for relationship inference. The node PR value is used to constrain the random walk range in the selection of the first step random walk link, which effectively solves the problem of graph embedding based on the current situation. Extensive experiments are performed on the real datasets of three different cities to acquire the performance of the proposed algorithm. The evaluations show that the new algorithm improves the accuracy of relationship prediction in the field of relationship inference.