Modeling the Temporal Dynamics of Social Rating Networks Using Bidirectional Effects of Social Relations and Rating Patterns

In this paper we first observe and analyze the temporal behavior of users in a social rating network on expressing ratings and creating social relations. Then, we model the temporal dynamics of a SRN based on our observations and using bidirectional effects of ratings and social relationships. While existing models for other types of social networks have captured some of the factors, our model is the first one to represent all four factors, i.e. social relations-on-ratings (social influence), social relations-on-social relations (transitivity), ratings-on-social relations (selection), and ratings-on-ratings (correlational influence). We also model the strength of each effect throughout the evolution of a SRN. Using our model, we develop a generative model for SRNs. Such a model can serve as basis for several purposes, in particular link prediction, rating prediction and prediction of future community structures. Given the sensitive nature of social network data, there are only very few public social rating network datasets. This motivates the development of generative models to create such synthetic datasets for research purposes. Our experimental study on the Epinions dataset demonstrates that the proposed model produces social rating networks that agree with real world data on a comprehensive set of evaluation criteria much better than existing models.