An Influence Field Perspective on Predicting User's Retweeting Behavior

Yi Shen, Jianjun Yu, Kejun Dong, Juan Zhao, Kai Nan

Abstract: User’s retweeting behavior, which is the key mechanism for information diffusion in the micro-blogging systems, has been widely employed as an important profile for personalized recommendation and many other tasks. Retweeting prediction is of great significance. In this paper, we believe that user’s retweeting behavior is synthetically caused by the influence from other users and the post. By analogy with the concept of electric field in physics, we propose a new conception named “influence field” which is able to incorporate different types of potential influence. Based on this conception, we provide a novel approach to predict user’s retweeting behavior. The experimental results demonstrate the effectiveness of our approach.