An enhanced trust propagation approach with expertise and homophily-based trust networks

Abstract

The transitivity property of trust enables the propagation of a trust value through a chain of trusting users in social networks and then provides an expected trust value for another user. Logically, a user in social networks can assess a large number of other users, even if two users have not been directly connected previously. However, a large percentage of trust propagation efforts fail to find reliable trust paths from a source user to a target user because the web of trust in real-world online social networks is too sparse. The success (both quality and quantity) of a trust propagation algorithm strongly relies on the density of a web of trust. The more trust paths that are able to reach the given target user, the more reliable will be the trust estimates based on the trust path with the highest strength. In this paper, we propose an enriched trust propagation approach by combining a homophily-based trust network with an expertise-based trust network, which enhances the density of the trust network. We then evaluate the prediction accuracy and coverage of trust propagation based on various aggregation methods and highlight the most promising method.