Predicting information diffusion via matrix factorization based model

Existing information diffusion models usually model the diffusion process based on the underlying networks, while the diffusion networks in real world are more complex than that of the underlying networks. In this paper, we propose a matrix factorization based predictive model (MFPM) to directly model the diffusion process we had observed and predict the information diffusion states in the future. Experiments on real world datasets suggest that our model outperforms the state-of-the-art information diffusion models for information diffusion prediction tasks.