Implicit Trust Propagation

The concept of trust plays key role in facilitating interactions and transactions in online systems. It is a major tool for handling information overload in online systems and filtering information by incorporating trust in recommender systems. Network of explicitly trusted users is very sparse. Implicit trust relations can be inferred based on user attributes and behavior. New trust relationships among users can be inferred using trust propagation properties which are not connected by explicit trust to solve the data sparsity problem. Here, we evaluate how effective is indirect implicit trust in predicting the explicit trust using different metrics. We show that indirect implicit trust has relatively higher correlation with rating similarity compared to direct implicit trust. We are able to infer more trust relations using indirect implicit trust as seen from higher recall and coverage values but the accuracy in predicting trust is low compared to direct implicit trust.