Attribute Inference Based on User Similarity and Random Walk

Abstract

With the wide application of social media, attribute inference based on user behavior has an important application value in personalized recommendation, psychological diagnosis and social investigation. Most of the current researches on attribute inference are to directly build a heterogeneous information network between users and attributes, and use the network structure and user behavior to infer attributes, while ignoring the similarity between users. In order to solve the above problems, this paper proposes an algorithm for attribute inference based on user similarity and random walk (USRW), which uses user-based collaborative filtering and random walk to calculate the score of target user for attributes that they do not own by using user similarity and network structure respectively, and then combines the two in a weighted manner for attribute inference. In this paper, the proposed algorithm is tested on the public dataset Deezer Social Networks and compared with random walk and user-based collaborative filtering algorithms. The experimental results show that attribute inference based on user similarity and random walk has higher inference performance than random walk and user-based collaborative filtering algorithms.