A social network-aware top-N recommender system using GPU
Proceedings of the 11th annual international ACM/IEEE joint conference on Digital libraries
| Hi-index | 0.00 |
Social networks are becoming more and more popular with the advent of numerous online social networking services. In this paper, we explore social rating networks, which record not only social relations but also user ratings for items. We analyze and model the effects of social influence and correlational influence in such networks, based on influence coefficients that measure the degree of influence in a network. We distinguish two types of user behavior: adopting an item and adopting a rating value for that item. We propose models to analyze and measure the influence of neighbors on both item and rating adoption behavior of users. Our experiments demonstrate that social influence has a much stronger impact on user behavior than correlational influence. Social and correlational influence are global effects in the entire network. However, there are local differences, i.e. certain users have a stronger social influence than others. To model this effect, we introduce the novel concept of social authority of individual users. We also propose an objective way to evaluate the social authority measure by injecting it into a simple recommender system.