Modeling Information Diffusion in Social Networks Using Latent Topic Information

In present scenario, social networking and microblogging sites have become dynamic and widely used media for communication. Here people share information on various topics through which they express their likes and interests. Analyzing the process of information diffusion on these platforms not only helps in understanding the underlying social dynamics, but is also important for various applications like marketing and advertising. In this paper, we explore a novel problem in social network analysis which is to identify the active edges in the diffusion of a message in the social network. We cast this task as a binary classification problem of detecting whether a link in the social network participates in the propagation of a given message. We propose a learning-based framework which uses user interests and content similarity modeled using latent topic information, along with the features related to the social network. We evaluate our model on data obtained from a well-known social network platform - Twitter. The experiments show a significant improvement over the existing methods.